Add tests & fixes for Code Fixes

[Booksbaum]

In this PR: Just "simple" CodeFix cases (-> not to many tests necessary, no huge issues or big necessary changes).

New CodeFixes & more complex cases (like GenerateInterfaceStub) get each an extra PR.

Code Fixes

• Add tests for CodeFixes:
• Rename some CodeFix modules & files to match their counterpart in dotnet/fsharp (VS)
  • Note: Not all names were adjusted. Some names I kept because I think their names are more descriptive.
  • Note: In listed fixes below: new names are used

• Fix: GenerateAbstractClassStub triggers twice
  • Note: Triggered for Errors 54 ('Type is abstract') and error 365 ('No implementation for this members').
    I think both errors always occur together
    -> Now just triggers for error 365
  • Note: CodeFix still has some other issues. See pending tests & commit msg (-> is a more complex case)
• Fix: AddMissingRecKeyword uses wrong function name in title
• Fix: AddTypeToIndeterminateValue doesn't trigger
• Fix: UseMutationWhenValueIsMutable adds <- after = instead of replacing it
• Fix: ConvertCSharpLambdaToFSharpLambda joins multi-line body to single-line body
• Fix: MakeDeclarationMutable doesn't trigger in untitled file
• Fix: AddMissingEqualsToTypeDefinition doesn't trigger
• Fix: AddMissingEqualsToTypeDefinition adds space on wrong side of =
  • Note: Was valid, just looked odd
  • Note: Wasn't visible because CodeFix didn't trigger
• Fix: ConvertPositionalDUToNamed doesn't put space before wildcard
  • Note: Was valid, just looked odd
• Change: Trigger ChangeDerefBangToValue on Diag 3370

  • prev: always check: cursor on ref expression that's accessed with !

  • now: trigger on diag 3370
    -> simpler (and most likely faster)

  • But change in behaviour:

    • prev: trigger on ref expression
    • now: trigger on bang (! -> location of diag)

    Example:

    let fr a = ref a
    let v = !(fr 5)
    //       ^^^^^^ prev trigger range
    //      ^ new trigger range

    I think that's ok: bang gets replaced -> trigger ON bang

• Fix: ResolveNamespace doesn't trigger when target is in last line
  • Note: Threw exception -> prevented other CodeFixes from running
• Fix: ReplaceWithSuggestion surrounds identifiers in double-backticks in additional double-backticks (-> 4 backticks: ````hello world````)
• Fix: RemoveUnusedOpens doesn't trigger in first line

Other Fixes

• Fix: Goto Declaration fails when declaration is in untitled file

---

Note: I only fixed what I stumbled across while adding tests.
There might or might not be any corresponding existing issues, which get fixed (or not fixed) in this PR.

---

Not handled (-> complex cases)

• ResolveNamespace
  • Needs unification with Auto-open via Code Completion (currently disjoint implementation and open in different location)
  • -> share (at least some) tests
• GenerateInterfaceStub
  • I'm currently working on this (-> use InterfaceStubGenerator in dotnet/fsharp instead of custom/copied one (I think was previously not public in dotnet/fsharp) )
• GenerateAbstractClassStub
• GenerateRecordStub
• GenerateUnionCases
  • I'm currently working on this

Note: There are some fixes for some of these CodeFixes,
but that's just something small and doesn't cover everything.

---

Next steps (already in progress):

• Add missing CodeFixes from dotnet/fsharp (-> Visual Studio)
• Update GenerateInterfaceStub
• Update GenerateUnionCases

Question that came up while toying with GenerateUnionCases: Is there a way in FCS to get the type of expression? It seems it's just possible for identifiers, but not expressions in general?

[baronfel]

I've got over it commit by commit (thank you for the nice separation, by the way!) and this is another wonderful chunk of work. Thank you for taking on this task! It's incredibly appreciated.

I have two small notes where we should be able to remove some of our patched/ported FCS functions now that we've updated, if you wouldn't mind doing that now, since we have the safety net of your new tests.

[baronfel]

@Booksbaum anything else you want to get into this? Otherwise I'll merge.

[Booksbaum]

No, all good