use indexmap instead of linked hashmap, switch hasher to use ahash

[jupyterkat]

ahash is straight up faster for string hashing
indexmap is at least maintained and has methods that we use, also faster iteration.

before:

after:

[SpaceManiac]

I saw a much smaller speedup on my machine, over several trials: dreamchecker average time from 5.946s to 5.794s, dm-langserver average time from 6.488s to 6.344s. Not nothing, but not as drastic as you reported. Can you try doing a few extra trials so we can find out if it your numbers were a fluke (disk cache can be a big confounder), or if there's a difference between our machines?

[jupyterkat]

might be because your machine just doesn’t support AES-NI like most modern machines do. have you tried on larger codebases where the extra seconds matter more

[SpaceManiac]

As far as I can tell my processor does support AES-NI. I was testing on /tg/station.