add regressions to changelog (NEWS.md)

[aplavin]

Following recent slack discussion, https://julialang.slack.com/archives/C67910KEH/p1683570942857959.

Brief summary of that:

Would it make sense to add known regressions to release notes (NEWS.md)? For 1.9 vs 1.8, I've experienced two performance regressions myself, and seen another one discussed:
<...>
Surely there are more, but how can users become aware of them?

Seems reasonable to me; I would want to know about known regressions in a release if I was considering whether to update or not

References issues: #48612 (mine) JuliaLang/LinearAlgebra.jl#998 (@pablosanjose) #48217 (@Moelf). Feel free to append more, or give more details!

[KristofferC]

The tone of these bullet points feels off to me and the descriptions are also much more general than the actual concrete issue. I think it is best to just link to the issues without trying to summarize them because any such attempt will lose too much details.

Something like:

• The following regressions have been identified: [#48612], [#49332]

[aplavin]

For a user considering updating his perfectly working code to a new julia version, regressions are as, or even more important than new features.
Features are described in NEWS inline, as short summaries with links to relevant issues/PRs - clearly, it's considered useful to have a summary at a glance. Should make sense to follow the same format for regressions.

Sorry if the tone sounds off, that was my crude attempt at making issue titles sound reasonable as bullet points. As a non-native speaker, I would appreciate any improvements here.

[giordano]

Quoting myself from slack:

Remember to add the improvement back in the release notes of 1.10 too

[aplavin]

I don't know how to make PR with changes to two branches at once

[aplavin]

bump...
What remains to be done on my side to add this?

[KristofferC]

The set of regressions in a release is a moving target since new ones can be found at any point. Therefore, I don't think it makes sense to store this information in a place that can not be mutated. How about adding a GitHub issue label (e.g. "regression 1.9") and linking to the list of such issues from this place?

[KristofferC]

Concretely: https://github.com/JuliaLang/julia/issues?q=+label%3A%22regression+1.9%22+

[aplavin]

Thanks @KristofferC, that's a nice suggestion! Implemented by adding a link to news.md.
What I don't understand is how to mark an issue as a regression. For example, I reported one more (#50130), but don't see how to add a label there.

[aplavin]

Gentle bump...

[aplavin]

@KristofferC so how to mark regression issues for them to appear in the list?
Found these unmarked ones from a cursory search:
#49935
#49241
#50073
#50130
#49545
#50110
#49756
#49697