Generate `*_FIRST` parser `TokenSet`s

[Veykril]

We should be able to mostly generate these from the ungrammar file I think instead of maintaining the manually (and risking them becoming outdated which they partially are),

[davidsemakula]

@rustbot claim

[Veykril]

@davidsemakula did you already put time into this specifically? If not I'd ask if I can claim this issue, because I technically already have an implementation for this that I just need to adapt for rust-analyzer otherwise. Wouldn't want you do to do work that's not technically necessary here :)

[davidsemakula]

@Veykril not significant time, you can re-claim it 🙂