Update preferInferredTypes rule to also support converting from inferred types to explicit types, fix bugs

[calda]

This PR renames the preferInferredTypes rule to propertyType, and makes the rule support both directions of conversions. It now fully respects the --redundantType inferred/explicit/infer-locals-only option, and can convert freely between

let foo: Foo = .init()

and

let foo = Foo()

  // --redundantType inferred
- let foo: Foo = .init()
+ let foo = Foo()

- let bar: Bar = .defaultValue
+ let bar = Bar.defaultValue

  // --redundantType explicit
- let foo = Foo()
+ let foo: Foo = .init()

- let bar = Bar.defaultValue
+ let bar: Bar = .defaultValue

  // --redundantType prefer-locals-only
  class Foo {
-     let foo = Foo()
+     let foo: Foo = .init()

      func bar() {
-         let bar: Bar = .defaultValue
+         let bar = Bar.defaultValue
      }
  }

That being said each option has various edge cases that may require manual fixes. When I ran these on Airbnb's app codebase:

• inferred had about 15-20 cases that I had to fix manually (not so bad considering we have like 2M+ LOC!)
• infer-locals-only mostly only had issues with cases where we had functions that look like types (e.g. func Translation(...)) or static members that don't return the same type. I added a --preservesymbols option that lets you exclude any sort of type or property that has this issue, which solved most of these issues. After this it was only miscellaneous issues that were individually easy to work around.
• I didn't try to use explicit, but it seems reasonable to support it if somebody wants to use it. I expect this would be somewhat difficult, but it's definitely possible.

I'd be fine with leaving the rule as-is (preferInferredTypes only supporting inferred rather than explicit or infer-locals-only) if you think the infer-locals-only and explicit options are too unreliable. I find inferred and infer-locals-only definitely reliably enough for use in Airbnb's codebase / style guide.

---

This PR also fixes other issues where the existing code would cause build failures:

For example, this compiles:

extension String {
  static let foo = "Foo"
}

let foo: String? = .foo

but this doesn't:

extension String {
  static let foo = "Foo"
}

// error: type 'String?' has no member 'foo'
let foo = String?.foo

I would have posted the two changes as separate PRs, but the propertyType rewrite depended on the bugfixes.

[calda]

The redundantInit rule didn't work correctly when run after the preferInferredType rule. At first I fixed this by changing the redundantInit rule to not check the operator type, but thinking about it more it seems better to have the preferInferredType rule change the "." operator from .prefix to .infix.

[calda]

Unrelated bugfix, this OptionDescriptor key path was wrong

[codecov]

Codecov Report

Attention: Patch coverage is 98.30508% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 95.17%. Comparing base (ca70962) to head (532eab8).

Files	Patch %	Lines
Sources/Rules.swift	98.31%	3 Missing ⚠️
Sources/ParsingHelpers.swift	97.95%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #1658      +/-   ##
===========================================
+ Coverage    95.13%   95.17%   +0.03%     
===========================================
  Files           20       20              
  Lines        22707    22880     +173     
===========================================
+ Hits         21603    21775     +172     
- Misses        1104     1105       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[calda]

We can't detect this unless the code uses the optional existential any syntax, so I added a note about it in the "known issues" section. Within the Airbnb codebase all existing examples of this pattern do use the any syntax, so this is ok for us.

[calda]

With the latest changes, you could add --preservesymbols ShapeStyle to avoid this problem when without the existential any syntax.

[calda]

@nicklockwood, I updated the rule implementation to support both prefer-locals-only and explicit rather than just inferred like it did before. This PR also includes a bunch of bug fixes. I added some context and commentary to the PR description, let me know what you think.

[calda]

We may also want to rename the --redundantType option to --propertyType and then have both the redundantType and propertyType rule respect it. It's slightly weird for the propertyType rule to use the --redundantType option, but I agree your thinking that we do want them to be configured using the same option rather than two separate options (so there's never a conflict). I'm also fine with leaving it as-is.

[nicklockwood]

@calda thanks so much for these features and fixes, I really appreciate it. Sorry I've not had very much bandwidth to discuss the design. I'm not certain yet what the best approach is with respect to naming. I'll try to collate my thoughts and get back to you in the next few days.

[calda]

Thanks, no worries :) appreciate the merges, we can iterate on develop