Tweak is_nonspacing_mark_name based on flamegraph

[rsheeter]

Before change is_nonspacing_mark reports as ~4% of total time and a major part of GlyphIrWork:

Essentially all the time was spent on partial eq for str so in this PR I try to avoid that by exiting early when the str could not possibly match.

After change it reports as ~1.5% and while still visible it no longer dominates GlyphIrWork:

Tested using Oswald.

[cmyr]

Okay so to check my understanding:

• we have, from fonttools, a list of the names of all possible nonspacing marks
• for all our glyph names, we need to check if they're on that list

I wonder what other approaches might work for us? One option would be a compile-time perfect hash function, using a crate such as https://docs.rs/phf/latest/phf/? Apparently this also has a codegen mode, so we wouldn't necessarily even need to bring it in as a dep.

It might be worth at least trying that, and seeing what the perf looks like?

[rsheeter]

I have deferred exploration of phf, which does seem potentially promising, to #399