Fix unstable formatting of trailing end-of-line comments of parenthesized attribute values

[MichaReiser]

Summary

Fixes #16151

This fixes a formatter instability where it incorrectly moved a trailing end-of-line comment
of a parenthesized attribute value was moved to the next line:

result = (
    (await query_the_thing(mypy_doesnt_understand))  # type: ignore[x]
    .foo()
    .bar()
)

Got reformatted to

result = (
    (await query_the_thing(mypy_doesnt_understand))
      # type: ignore[x]
    .foo()
    .bar()
)

and then finally to:

result = (
    (await query_the_thing(mypy_doesnt_understand))
    # type: ignore[x]
    .foo()
    .bar()
)

Not only is it an instability, but it also moves the type: ignore comment away from the node it was attributing.

This PR fixes both by preserving the input formatting. Both those examples are correct formattings now:

result = (
    (await query_the_thing(mypy_doesnt_understand))  # type: ignore[x]
    .foo()
    .bar()
)


```py
result = (
    (await query_the_thing(mypy_doesnt_understand))  # type: ignore[x]
    .foo()
    .bar()
)

result = (
    (
        await query_the_thing(mypy_doesnt_understand)  # type: ignore[x]
    )  # other
    .foo()
    .bar()
)

I don't expect this to change any stable formatting because the formatter always moved trailing end-of-line comments and made them leading own-line comments of the . token

Test Plan

Added tests, ecosystem check

[github-actions]

ruff-ecosystem results

Formatter (stable)

✅ ecosystem check detected no format changes.

Formatter (preview)

✅ ecosystem check detected no format changes.

[huonw]

Thanks for fixing this! (BTW, I think the code examples in the PR description aren't quite the right ones.) 😄

[MichaReiser]

@huonw thanks for making me aware. I was in the middle of writing the summary and then realized that the fix wasn't ideal. So I decided to push what I had.

[MichaReiser]

We want to fall through to the next case if the comment is after the closing parentheses

[dhruvmanila]

I'm a bit unfamiliar with fluent chain layout so it might good to get a second pair of eyes on this (Konsti?) but otherwise looks good.

[dhruvmanila]

What's the significance of this test case? It seems to be working fine on main

[MichaReiser]

I mainly wanted to make sure that I don't break this case.

[MichaReiser]

I'm a bit unfamiliar with fluent chain layout so it might good to get a second pair of eyes on this (Konsti?) but otherwise looks good.

Yeah. I don't think anyone is very familiar with this specific code at the moment :)