automata/nfa/thompson: tweak allocation handling for UTF-8 compiler

[BurntSushi]

This commit makes the UTF-8 compiler allocation pattern a bit more lazy.
Instead of allocating 10,000 entries up-front, we now allocate via a
smaller chunking pattern. That is, we only allocate N chunks up front,
and each chunk is itself only allocated only when we actually try to put
something inside of it.

This also adds a few unit tests covering existing and new behavior.

This addresses #1316 by making compilation of that regex substantially
faster. Previously, this program

fn main() {
    for _ in 0..100000 {
        let re = regex::bytes::RegexBuilder::new(r"^[^/]+/foo/[^/]+$")
            .unicode(true)
            .build()
            .unwrap();
        assert!(re.is_match(b"bar/foo/baz"));
    }
}

took about 1.86s to run. But with this change, it only takes 1.14s to
run. This still isn't quite as good as the non-Unicode case (which takes
0.64s), but we're closer with this change.

Fixes #1316