build, meta: drop Python 3.7 support, checks, and CI job

[dgw]

Description

This fulfills the last "Finally letting the dead snakes lie." part of the 8.0.0 milestone's description. Python 3.7 was officially EOL'd as of a month ago, on June 27.

Checklist

• [x] I have read CONTRIBUTING.md
• [x] I can and do license this contribution under the EFLv2
• [x] No issues are reported by make qa (runs make quality and make test)
  • (As of when I prepped the patch.)
• [x] I have tested the functionality of the things this change touches
  • When I wrote it, I did, with make test etc. (Now it needs a rebase for retesting.)

Conflict notes

@SnoopJ and @Exirel arrived at similar conclusions for dealing with aliases in plugins/rules.py; this has a conflict with 098d8c309058dc794dd04d0d464598482ca94505, but the approaches are quite similar. That and b0693b7b0da1047aa0f509a30f9d15ef6b71065d are both from #2471, which touched a bunch of stuff (so I knew the risk of conflicts with this relatively long-lived housekeeping branch was inevitable).

The only other little conflict is more directly my fault. While I set @SnoopJ loose on my todo list for #2471 to get it done faster, #2480—specifically, 85f8b82d3917135e309b6ce4efc601d4c79560a7—is all me. 🙈

Guess I have to rebase this one, huh? Damn, I love merging branches that diverged long before master because the history graph looks cool, but sometimes it's just not meant to be. 😅

Special thanks

…to the other maintainers for humoring me when I said I wanted to reach #2500 and make this PR a numerical milestone as well as a development one. You're all awesome for tossing in extra issues & PRs to get us here! ❤️

[dgw]

• [x] No issues are reported by make qa (runs make quality and make test)
  • (As of when I prepped the patch.)

Well, now the latest version of pyflakes (updated from 3.0.1 to 3.1.0) raises "F811 Redefinition of unused name" for:

https://github.com/sopel-irc/sopel/blob/e63afba333ea47ca02248e83b495544bc8b256f0/sopel/plugins/rules.py#L1782

The name execute_handler is defined earlier, but never used:

https://github.com/sopel-irc/sopel/blob/e63afba333ea47ca02248e83b495544bc8b256f0/sopel/plugins/rules.py#L1766

Upstream seems to consider this as "Working as intended" (PyCQA/pyflakes#780), so I guess we'll have to if: ... else: this one now. 🙁

[dgw]

Having whittled down most of the open threads from @SnoopJ and myself, I figured it's worth seeing if Exi has free time later this week(end) to chime in.

[SnoopJ]

• I think importlib_metadata can be removed from a few more places, unless 3.8 doesn't have importlib.metadata.EntryPoint()

It does have EntryPoint but I believe the behavior of entry_points() changes subtly between 3.9 and 3.10 which keeps the dependency around? I don't know the details myself, but see this comment

[dgw]

• modules/tld.py has a "py<3.5" comment

My pattern missed that because of the <, whoops. Thanks.

[dgw]

I decided to throw replacing importlib_metadata usage under test/ at the CI for fun. Nothing broke on 3.8/3.9, so 🚀 I guess.