Optimize comparisons to chars

[miniBill]

[jfmengels]

This is super weird, because I can reproduce tests being many time faster, but when I run the performance script

./find-regressions.sh --no-check ~/.elm/0.19.1/packages/

then performance is about the same, maybe even worse

Published: avg 1.5168321927699087
Current:   avg 1.5336765493284654
Diff: -1.0982991534904896% faster

[jfmengels]

Maybe it's time we extract this to a named function 😅

[lue-bird]

Using the code in benchmark/ I also don't notice any improvement.
So I imagine this difference in the test may be caused by our existing code not getting optimized outside of eol2 (== literal vs case).

In any case we should merge this.

[jfmengels]

Right, I agree I think it's a difference between having --optimize and not having it. And this change manually gets us closer to that. It's impressive that these tiny changes have so much impact on performance in non-optimized though.

Even if the final performance does not get improved (which it might, just a bit hard to measure), it makes tests run much faster, and that's a nice win for the people working on this project!

Thank you @miniBill ❤️

[jfmengels]

@lue-bird do you think this is worth releasing? 🤔

[jfmengels]

Published in 7.3.8!
This is also going to make elm-review faster when run with --debug, as well as elm-review rule tests ❤️

[lue-bird]

I applied the same changes for elm-syntax-format's lenient parser and got a 5-10% improvement (with eol2 applied) and 3.x for non eol2. That's amazing.