Add lrlex %expect-missing-tokens directive.

[ratmice]

This adds a lrlex directive %expect-missing-tokens for the docs PR #649

The syntax is like: %expect-missing-tokens "TOK" including quotes.

This is so that it behaves exactly like %expect-unused which accepts quoted tokens, but unquoted rules. Unlike %expect-unused, the new directive doesn't take unquoted rules as those are a parser thing.

I tested that the new test fails if we remove the new directive.

[ltratt]

Quick thought: would it be ambiguous if we shortened this to %expect-missing to match %expect-unused?

[ratmice]

I don't think it would be ambiguous, at worst we might add %expect-missing to lrpar someday, and grep might match both. But that said I'm not sure what justification there is for having the parser not handle a token the lexer is capable of producing. But that grepability is kind of spurred me to throw tokens on the end. I'm happy to remove it though.

[ratmice]

I went ahead and used the shorter name in 80a6b98