Add send-with-block benchmark

[tekknolagi]

Benchmarks send-with-block, especially polymorphic send-with-block. Gives us information on how much inlining-driven monomorphization is helping.

[k0kubun]

How long does each iteration take? My understanding is that the shorter each iteration is, the noisier the benchmark is, because it needs to go through non-benchmark code a lot (after each iteration). So if it takes 1000 iterations to run for the default benchmark duration, I would make < 10 larger.

However, if the intention is to encourage a JIT to inline everything from the block given to run_benchmark and avoid staying in iterations that do not compile the block too long, I'm fine with configuring it this way.

[tekknolagi]

plum% ruby benchmarks/block_methods.rb
ruby 4.1.0dev (2026-09-17T15:01:37Z mb-condbranch-chec.. c4ea80f56a) +PRISM [arm64-darwin25]
itr:   time
 #1:  359ms
 #2:  361ms
 #3:  369ms
 #4:  361ms
 #5:  365ms
 #6:  360ms
 #7:  363ms
^Cbenchmarks/block_methods.rb:44:in 'block in Object#kernel_tap': Interrupt
        from benchmarks/block_methods.rb:42:in 'Kernel#tap'
        from benchmarks/block_methods.rb:42:in 'Object#kernel_tap'
        from benchmarks/block_methods.rb:75:in 'block in <main>'
        from /Users/emacs/Documents/code/yjit-bench/harness/harness.rb:30:in 'Object#realtime'
        from /Users/emacs/Documents/code/yjit-bench/harness/harness.rb:56:in 'Object#run_benchmark'
        from benchmarks/block_methods.rb:68:in '<main>'

plum% ruby --zjit benchmarks/block_methods.rb
ruby 4.1.0dev (2026-09-17T15:01:37Z mb-condbranch-chec.. c4ea80f56a) +ZJIT stats +PRISM [arm64-darwin25]
itr:   time
 #1:  267ms
 #2:  261ms
 #3:  222ms
 #4:   84ms
 #5:   85ms
 #6:   86ms
 #7:   86ms
 #8:   84ms
 #9:   85ms
#10:   84ms
#11:   83ms
#12:   84ms
#13:   85ms
#14:   84ms
#15:   83ms
#16:   95ms
#17:   91ms
^Cbenchmarks/block_methods.rb:15:in 'block in Object#array_map': Interrupt
        from benchmarks/block_methods.rb:13:in 'Array#map'
        from benchmarks/block_methods.rb:13:in 'Object#array_map'
        from benchmarks/block_methods.rb:72:in 'block in <main>'
        from /Users/emacs/Documents/code/yjit-bench/harness/harness.rb:30:in 'Object#realtime'
        from /Users/emacs/Documents/code/yjit-bench/harness/harness.rb:56:in 'Object#run_benchmark'
        from benchmarks/block_methods.rb:68:in '<main>'

plum%