Language server improvements #12

#9 added basic support for Microsoft's language server protocol, and integration with Emacs and VS Code. However, there's a lot of improvements which could be made:

Features

• Code folding: Aside from the obvious indentation/structrual folding, we should allow folding multi-line doc comments into a single one. Ideally we'd also identifiy multiple local x = require(y) statements and allow folding them together.

• Refactoring/code fixes: Effectively we want a similar scheme to linters, just without detailed messages and tags. Some basic things we might like:

  • Invert if/else statement.
  • Convert between quote styles (single, double, long-style).
  • Convert between number formats (decimal <-> hex).

• Renaming: This should be conceptually quite simple - just generate an edit which updates all names. We need to be careful that we don't introduce a conflict though - possibly worth seeing what additional info the resolver needs to handle this efficiently.

• Document outline: I'm not entirely sure how we want to do this yet - we obviously don't want to document every local, but still need to give a decent overview of the layout.

• Workspace symbols: It should be sufficient just to load all modules and dump the entire symbol tree.

• Completion: This is an obvious choice, but I have no clue where to even begin on this.

• Handle references in doc comments: We should be able to use go-to-definition, see usages and have symbol highlights for references within doc comments.

Performance

• Check for memory leaks: It might be possible to do some automated detection of this too. For instance, run a specific task (open/close a file) many times and compare memory usage over time.
• Use LSP's file watching support

Editors

• Download the language server: On compatible platforms, we should download the language server automatically. Possibly worth looking into periodic updates too.