Make errors nicer when TDNR fails

[runarorama]

Fixes #4632

This makes the error message for TDNR failures simpler and clearer.

• Removes references to the namespace, since we're usually in the project root and UCM tells you which namespace you're in anyway.
• Infers the likely cause a little better and doesn't suggest e.g. a totally missing symbol when there are TDNR suggestions available.
• Added "TDNR masking the true error" to the list of common causes.
• Removed defunct instructions for adding a library.
• Added clear instructions for resolving the error
• Suffixified the TDNR suggestions.
• Removed bullets from TDNR suggestion list.
• Improved wrapping.
• Sorted suggestions by the length of the FQN, as a proxy for relevance.

Example of new behaviour:

[ChrisPenner]

Looks good! I'm curious what's up with the transcripts outputting _ as the expected type though 🤔

[aryairani]

Currently a compile error in parser-typechecker/tests/Unison/Test/UnisonSources.hs

[aryairani]

Looks good! I'm curious what's up with the transcripts outputting _ as the expected type though 🤔

@runarorama I have this question too — It looks like it used to say

There are no constraints on its type.

but now it says

Its type should conform to _

(and missing the .)

Otherwise, I'm eager to get it merged before transcripts start failing again 😅

[runarorama]

Its type should conform to _ is a regression in trunk unrelated to this change.

I removed the . after the type because Foo.Bar -> Baz.Qux reads better than Foo.Bar -> Baz.Qux.

[aryairani]

Its type should conform to _ is a regression in trunk unrelated to this change.

Hm, but I don't see that output in the transcripts on trunk?

https://github.com/unisonweb/unison/blob/trunk/unison-src/transcripts/resolution-failures.output.md?plain=1#L127

There are no constraints on its type.

vs

https://github.com/unisonweb/unison/blob/runarorama/improve-tdnr-message/unison-src/transcripts/resolution-failures.output.md?plain=1#L121

Its type should conform to:
  
      _

I think it might be due to losing this logic, which is present here, but missing here and here.

I removed the . after the type because Foo.Bar -> Baz.Qux reads better than Foo.Bar -> Baz.Qux.

Ok 👍

[runarorama]

OK, I believe I have addressed all of the issues and fixed the places where it was missing the existentials check. Output looks good in all transcripts. 🚢

[aryairani]

Great!