DOC/DEV: update commit message guidance about scipy

[rgommers]

Agree with points 2 and 3. One comment on 2" backticks are not necessarily bad, and useful especially when referring to names with underscores, because that prevents formatting in release notes from going wrong (PR titles are included, and underscores are special syntax in reST.

Point 1 is a bit more complex. I agree with the first part. However, if there's a single commit that fixes an issue and you don't force-push that a lot, then adding Closes gh-1234 in the commit message is helpful. Commit messages need to be descriptive, and for future debugging it's quite relevant to see what a commit fixed without having to figure out what PR it was included in. So this is more a "it depends".

[ilayn]

For item 2, spaces are not needed around colons since commit message is limited space so better spend it on making the commit message clearer instead. Hence MAINT:fft:Some commit messageis fine and makes grepping much easier.

[lucascolley]

For item 2, spaces are not needed around colons since commit message is limited space so better spend it on making the commit message clearer instead. Hence MAINT:fft:Some commit messageis fine and makes grepping much easier.

Point taken, but most people do put spaces around the colons¹, so I think it's probably not worth enforcing either way - maybe saying that either is okay (or including both in examples) makes sense.

Footnotes

1. and I think it looks nicer... ↩

[lucascolley]

That all makes sense. This issue was partly motivated by the guidance to 'name and describe your PR as you would write a
commit message'. I think Closes gh-1234 is one case where it would be useful to draw a distinction between the two - "it depends" for commit messages, as you say, but almost never good for PR titles.