Implement code action to add module to cabal file

[VeryMilkyJoe]

Related to #3695.
Closes #3595.

When a cradle error occurs that the current module is unknown, we now offer code actions for each possible field to add the module to the most likely cabal file.

This PR is split into three commits which group the changes into:

• changes to ghcide;
• a cabal file parser package which will be moved out of hls in the future;
• and changes to the cabal-plugin.

When a cradle error is found to be related to an unknown module, we edit the error by adding information (the module's path, the most likely cabal file to add the module to) such that the cabal plugin can offer code actions for adding the module to a cabal file.
For each other-module or exposed-module field in a stanza, we check whether any source directory of the stanza contains the module. If this is the case, we construct a code action which adds the module to the cabal file, while adhering to the cabal file's previous formatting as best as possible.

[michaelpj]

Lots going on here, just a couple of high-level comments. I'll try to review it more later.

[michaelpj]

I'd probably just make this an internal library of the cabal plugin for now. We can split it out later, but additional packages do mean we have to do more releases etc.

[michaelpj]

Does this handle whitespace okay? I've had that problem with doing this before: since lexer-less parsing encourages you to consume whitespace as part of your syntactic productions, you end up with the source span including the whitespace.

[michaelpj]

I actually wonder whether this parser would be easier to write with a lexer pass. I think megaparsec supports that...

[uhbif19]

Why cabal parser implemented by cabal would not work for this task?

https://github.com/haskell/cabal/tree/master/Cabal-syntax

[fendor]

It would work, given the existence https://github.com/Bodigrim/cabal-add/. However, this was deemed a somewhat interesting idea to add a cabal exactprinter, but it is likely not the best way forward.

[Bodigrim]

I'm not sure that reimplementing Cabal parser is a sustainable approach.

https://github.com/Bodigrim/cabal-add/ inserts build-depends largely re-using Cabal-syntax, and probably inserting modules could be done in a similar fashion.

[fendor]

@Bodigrim Yeah, the cabal-add approach feels more viable than the custom parser.

Independently, I believe we should have a third-party parser for cabal files.