internal: Use rustc-hash in more places

[lnicola]

I think there's still a couple a lot of these left, some in tests and the proc macro server.

[Veykril]

Ideally we should add a clippy.toml and the disallowed-types lint with that set up to forbid using plain HashMap, now that lints in Cargo.toml are a thing

[lnicola]

Good idea, I forgot that was a thing. It feels like the hardest part is bikeshedding them, e.g. what should we use in lsp-server?

[Veykril]

Hmm, I guess lsp-server is the exception, it should be fine to use in there?

[lnicola]

It also doesn't seem to work very well with derive macros. Anyway, I'll give it a try.

[Veykril]

what derive macros we use emit hashmap related things? 🤔 or was that more a general note, not specific to r-a's codebase

[lnicola]

We have some hashmaps in lsp_ext and i think we have to move them into a submodule to allow the lint there, because they derive Serialize.

[Veykril]

Might as well just replace those with FxHashMap as well since we have the dependency there.

[lnicola]

I don't remember exactly why that didn't work, they might end up in lsp-types or something like that.

[Veykril]

I see, well excluding that module should be fine anyways.

[bors]

☔ The latest upstream changes (presumably #16394) made this pull request unmergeable. Please resolve the merge conflicts.

[Veykril]

#16470 supersedes this