ZJIT: Optimize `String#match?`

[tekknolagi]

This makes a Regexp and does an uncached call from C to match on that object. We can move at least some of this to Ruby probably.

[nirvdrum]

TruffleRuby already does a good chunk of this is in Ruby. It's done in a few different steps:

• String#match? coerces the supplied pattern into a Regexp and then makes the Regexp#match? call. It makes use of two primitives: one for a faster type check and one for the type coercion.
• Regexp#match? delegates to an internal helper class called Truffle::RegexpOperations. I'm not sure why it's done this way. I think some of it is a holdover from before primitives were added, but it might for better call site optimization, too. The helper class is so the runtime support code doesn't leak new methods into Regexp.
• Truffle::RegexpOperations#match? does the bulk of the work. There's a Symbol check on the str arg, but that's only because this is ultimately shared with Symbol#match?as well. It takes care of bounding the position argument and remapping it if the value is negative. That leads to a trivial check that is known to always be false. Then it maps the character position to byte position because that's what's needed by the regex engine. And, naturally, then there's a primitive to call into the regex engine.

I think a fair chunk of that could be used almost directly.