fix(rome_js_formatter): regression of call arguments

[ematipico]

Summary

This PR fixed #2720

The new logic checks if any arguments breaks and if so, it does a custom printing.

Doing so, it triggered a regression of the new code that we didn't see before:

This snippet:

test.expect(t => {
	t.true(a);
}, false /* inline comment */ );

Would be formatted

test.expect ((t) => {
	t.true(a);
}, false /* inline comment */ );

Note the space added before the first (. This regression is caused by a bug inside the function write_space_between_comment_and_token, which triggers a false positive condition and it adds a space before the token.

I will have to look into it later. In order to avoid to commit wrong code, I changed the snippet that was causing the false positive. Tracked in #2814

Another regression that is triggered by our printer is already tracked in #2815 . This issue is caused by how best_fitting! prints the second alternative.

I would like to shut down this reformat issue, because the regression that is fixed in this PR is actually more important then the format issue, I think.

Test Plan

cargo test and checked that the regression is not fixed

PR
File Based Average Prettier Similarity: 77.44%
Line Based Average Prettier Similarity: 72.72%

main
File Based Average Prettier Similarity: 77.33%
Line Based Average Prettier Similarity: 72.61%

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Pages

Latest commit:	6a92176
Status:	✅  Deploy successful!
Preview URL:	https://8627c3b8.tools-8rn.pages.dev
Branch Preview URL:	https://fix-will-break-call-argument.tools-8rn.pages.dev

View logs

[github-actions]

Playground for commit 3b7e72f

[leops]

If I remember correctly calling .any will mutate the arguments_break iterator in place, so all the iterators items consumed to initialize any_breaks will not be processed by the next call to .any to initialize an_argument_breaks, is this really what's intended here ?

[ematipico]

That's really a good catch, and I was wondering why I could not replicate a reformat issue that I was encountering some time ago.