migrate typechecker to new lsp

[psteinroe]

migrates the typechecker to the new infrastructure.

• I did not introduce caching yet (despite it being possible) because it's fast enough imo.
• using run_async requires static lifetimes, so we need to clone all params - me not likey likey but fixing this probably requires a larger refactor. UPDATE: this is improved by running the typechecker in a single run_async :)
• I think we should break up the pull_diagnostics method and run one part async and another sync. UPDATE: I nerdsniped myself into doing the refactor. dont know if it actually makes a difference but it was a nice learning :)
• I did not spend much time implementing the adviser visitor trait for the typechecker diagnostics. i think we can keep it like that and modify it when feedback arrives. we could also think about parsing the pg error in specific diagnostic categories, e.g. typecheck/missingColumn etc but that requires a bit of play to find out about the different error types.
• any diagnostics other than the main message does not appear yet in the lsp output. need to check if its supposed to work.
• previously I used the custom parsers result to know where exactly the issue is. I switched this to tree sitter now and it works similarly well. At this point I am wondering why I even worked on the parser for half a year when we barely even need it 🥲🥲

note: I found there is still a crash in the change handler during testing. it happens when I add one or more newlines at the end of the file and then remove it again. will take a look this week. :)

[juleswritescode]

Awesome!!