Improve playwright comments

[djahandarie]

In #1676 we merged a change that completely broke all popups.

Our E2E testing (playwright) did detect the issue, but its output was ignored.

My analysis is that it was ignored because the output of playwright-report-summary:

1. does not make a clear visual differentiation between flaky tests and failures, requiring the user to carefully analyze the output
2. does not link to the report artifact, making it high friction for the user to actually check the report to see what changed
3. does not have a clear call to action instructing the user what they should do

I believe it is better to have our own custom comments (basically revert #1543 and improve the original comments a bit to properly differentiate between flaky and failure). That is what this PR does.

This achieves:

• Very small comment size, with just a single icon, making it easy to read and interpret
• Clear call to action to user instructing them what to do based on results
• Low friction via a direct link to the playwright-report (made possible by the new v4 artifact system on GitHub)
• Proper differentiate between flaky and failure, choosing to interpret flaky as ✔️ (but still comment to instruct user to improve tests) and failure as ⚠️ so it's hard to miss.

[djahandarie]

Example result: djahandarie#2 (comment)