Add optimized SHA1 and SHA256 using ARMv8 Crypto Intrinsics.

[samyron]

Add optimized SHA1 and SHA256 using ARMv8 Crypto Intrinsics. These instructions will be detected at runtime on MacOS, Linux and Android using the appropriate operating system specific mechanism.

Note: These implementations came from https://github.com/noloader/SHA-Intrinsics. This PR only includes the ARMv8 implementations as I do not have an x86 CPU with the correct instruction set support for testing those implementations.

These should recover some of the speed after removing OpenSSL support mentioned in issue #35 on ARMv8.

This commit adds the following extconf.rb flags:

1. --without-common-digest - Disable CommonCrypto detection to use one of the bundled implementations.
2. --without-arm-crypto - Disable ARMv8 Crypto extension detection to force the portable C fallback implementation.

This commit also includes NIST test vectors for SHA1 and SHA256 to verify the compiled implementation works correctly.

These were executed on an M1 Macbook Air.

=== SHA1 / short (16 bytes) -- backend: Portable C ===
ruby 3.4.8 (2025-12-17 revision 995b59f666) +PRISM [arm64-darwin24]
Warming up --------------------------------------
     SHA1 Portable C   200.019k i/100ms
Calculating -------------------------------------
     SHA1 Portable C      1.986M (± 1.3%) i/s  (503.47 ns/i) -      6.001M in   3.021090s

Comparison:
  SHA1 ARM Crypto:  2214084.8 i/s
  SHA1 Portable C:  1986226.8 i/s - 1.11x  slower
SHA1 CommonCrypto:  1350872.5 i/s - 1.64x  slower

=== SHA1 / medium (1024 bytes) -- backend: Portable C ===
ruby 3.4.8 (2025-12-17 revision 995b59f666) +PRISM [arm64-darwin24]
Warming up --------------------------------------
     SHA1 Portable C    55.810k i/100ms
Calculating -------------------------------------
     SHA1 Portable C    550.800k (± 4.2%) i/s    (1.82 μs/i) -      1.674M in   3.039760s

Comparison:
  SHA1 ARM Crypto:  1056399.8 i/s
SHA1 CommonCrypto:   832120.4 i/s - 1.27x  slower
  SHA1 Portable C:   550800.1 i/s - 1.92x  slower

=== SHA1 / large (1048576 bytes) -- backend: Portable C ===
ruby 3.4.8 (2025-12-17 revision 995b59f666) +PRISM [arm64-darwin24]
Warming up --------------------------------------
     SHA1 Portable C    77.000 i/100ms
Calculating -------------------------------------
     SHA1 Portable C    776.174 (± 0.5%) i/s    (1.29 ms/i) -      2.387k in   3.075341s

Comparison:
SHA1 CommonCrypto:     2307.0 i/s
  SHA1 ARM Crypto:     2058.3 i/s - 1.12x  slower
  SHA1 Portable C:      776.2 i/s - 2.97x  slower

=== SHA256 / short (16 bytes) -- backend: Portable C ===
ruby 3.4.8 (2025-12-17 revision 995b59f666) +PRISM [arm64-darwin24]
Warming up --------------------------------------
   SHA256 Portable C   187.472k i/100ms
Calculating -------------------------------------
   SHA256 Portable C      1.879M (± 0.3%) i/s  (532.16 ns/i) -      5.812M in   3.092709s

Comparison:
  SHA256 ARM Crypto:  2821313.8 i/s
SHA256 CommonCrypto:  2366863.2 i/s - 1.19x  slower
  SHA256 Portable C:  1879139.6 i/s - 1.50x  slower

=== SHA256 / medium (1024 bytes) -- backend: Portable C ===
ruby 3.4.8 (2025-12-17 revision 995b59f666) +PRISM [arm64-darwin24]
Warming up --------------------------------------
   SHA256 Portable C    25.901k i/100ms
Calculating -------------------------------------
   SHA256 Portable C    258.693k (± 0.3%) i/s    (3.87 μs/i) -    777.030k in   3.003673s

Comparison:
SHA256 CommonCrypto:  1217067.5 i/s
  SHA256 ARM Crypto:  1187297.8 i/s - 1.03x  slower
  SHA256 Portable C:   258693.3 i/s - 4.70x  slower

=== SHA256 / large (1048576 bytes) -- backend: Portable C ===
ruby 3.4.8 (2025-12-17 revision 995b59f666) +PRISM [arm64-darwin24]
Warming up --------------------------------------
   SHA256 Portable C    29.000 i/100ms
Calculating -------------------------------------
   SHA256 Portable C    288.299 (± 4.9%) i/s    (3.47 ms/i) -    870.000 in   3.017705s

Comparison:
SHA256 CommonCrypto:     2330.5 i/s
  SHA256 ARM Crypto:     1889.2 i/s - 1.23x  slower
  SHA256 Portable C:      288.3 i/s - 8.08x  slower

Additional benchmarks running Debian on Virtualbox on the same M1 Macbook Air

=== SHA1 / short (16 bytes) -- backend: Portable C ===
ruby 3.4.1 (2024-12-25 revision 48d4efcb85) +PRISM [aarch64-linux]
Warming up --------------------------------------
     SHA1 Portable C   158.975k i/100ms
Calculating -------------------------------------
     SHA1 Portable C      1.634M (± 1.6%) i/s  (612.06 ns/i) -      4.928M in   3.016349s

Comparison:
SHA1 ARM Crypto:  1663326.0 i/s
SHA1 Portable C:  1633837.8 i/s - same-ish: difference falls within error

=== SHA1 / medium (1024 bytes) -- backend: Portable C ===
ruby 3.4.1 (2024-12-25 revision 48d4efcb85) +PRISM [aarch64-linux]
Warming up --------------------------------------
     SHA1 Portable C    46.630k i/100ms
Calculating -------------------------------------
     SHA1 Portable C    467.343k (± 0.4%) i/s    (2.14 μs/i) -      1.446M in   3.093081s

Comparison:
SHA1 ARM Crypto:   916649.5 i/s
SHA1 Portable C:   467343.1 i/s - 1.96x  slower

=== SHA1 / large (1048576 bytes) -- backend: Portable C ===
ruby 3.4.1 (2024-12-25 revision 48d4efcb85) +PRISM [aarch64-linux]
Warming up --------------------------------------
     SHA1 Portable C    65.000 i/100ms
Calculating -------------------------------------
     SHA1 Portable C    656.318 (± 0.5%) i/s    (1.52 ms/i) -      2.015k in   3.070158s

Comparison:
SHA1 ARM Crypto:     2066.3 i/s
SHA1 Portable C:      656.3 i/s - 3.15x  slower

=== SHA256 / short (16 bytes) -- backend: Portable C ===
ruby 3.4.1 (2024-12-25 revision 48d4efcb85) +PRISM [aarch64-linux]
Warming up --------------------------------------
   SHA256 Portable C   165.590k i/100ms
Calculating -------------------------------------
   SHA256 Portable C      1.655M (± 0.5%) i/s  (604.36 ns/i) -      4.968M in   3.002281s

Comparison:
SHA256 ARM Crypto:  2118511.1 i/s
SHA256 Portable C:  1654641.7 i/s - 1.28x  slower

=== SHA256 / medium (1024 bytes) -- backend: Portable C ===
ruby 3.4.1 (2024-12-25 revision 48d4efcb85) +PRISM [aarch64-linux]
Warming up --------------------------------------
   SHA256 Portable C    22.921k i/100ms
Calculating -------------------------------------
   SHA256 Portable C    230.231k (± 0.4%) i/s    (4.34 μs/i) -    710.551k in   3.086248s

Comparison:
SHA256 ARM Crypto:  1057810.6 i/s
SHA256 Portable C:   230231.4 i/s - 4.59x  slower

=== SHA256 / large (1048576 bytes) -- backend: Portable C ===
ruby 3.4.1 (2024-12-25 revision 48d4efcb85) +PRISM [aarch64-linux]
Warming up --------------------------------------
   SHA256 Portable C    26.000 i/100ms
Calculating -------------------------------------
   SHA256 Portable C    260.243 (± 0.4%) i/s    (3.84 ms/i) -    806.000 in   3.097108s

Comparison:
SHA256 ARM Crypto:     1848.7 i/s
SHA256 Portable C:      260.2 i/s - 7.10x  slower

Additional benchmarks including OpenSSL

ruby 3.4.8 (2025-12-17 revision 995b59f666) +PRISM [arm64-darwin24]
Warming up --------------------------------------
   SHA256 Portable C    25.861k i/100ms
      SHA256 OpenSSL    76.391k i/100ms
Calculating -------------------------------------
   SHA256 Portable C    254.553k (± 2.3%) i/s    (3.93 μs/i) -    775.830k in   3.047817s
      SHA256 OpenSSL    750.597k (± 2.2%) i/s    (1.33 μs/i) -      2.292M in   3.053211s

Comparison:
SHA256 CommonCrypto:  1209718.4 i/s
  SHA256 ARM Crypto:  1169577.7 i/s - same-ish: difference falls within error
     SHA256 OpenSSL:   750596.7 i/s - 1.61x  slower
  SHA256 Portable C:   254552.7 i/s - 4.75x  slower


=== SHA256 / large (1048576 bytes) -- backend: Portable C ===
ruby 3.4.8 (2025-12-17 revision 995b59f666) +PRISM [arm64-darwin24]
Warming up --------------------------------------
   SHA256 Portable C    29.000 i/100ms
      SHA256 OpenSSL   233.000 i/100ms
Calculating -------------------------------------
   SHA256 Portable C    290.571 (± 0.7%) i/s    (3.44 ms/i) -    899.000 in   3.093913s
      SHA256 OpenSSL      2.329k (± 0.8%) i/s  (429.36 μs/i) -      6.990k in   3.001219s

Comparison:
SHA256 CommonCrypto:     2332.4 i/s
     SHA256 OpenSSL:     2329.1 i/s - same-ish: difference falls within error
  SHA256 ARM Crypto:     1872.5 i/s - 1.25x  slower
  SHA256 Portable C:      290.6 i/s - 8.03x  slower