irc: properly manage exception of the run-forever loop

[Exirel]

Description

Tin. Based on #2430, closes #2350.

To elaborate a bit: I realized that having a connection error and other uncaught exception in the run-forever loop (await read task, await write task) would lock the bot, and so it's not just friendly error messages, it is also a better exception handling.

Checklist

• I have read CONTRIBUTING.md
• I can and do license this contribution under the EFLv2
• No issues are reported by make qa (runs make quality and make test)
• I have tested the functionality of the things this change touches

[dgw]

I'm nitpicking… Maybe the final commit/PR title line should say "exceptions in the run-forever loop"?

[dgw]

Is this in a state I could check out on one of my bots for testing? Seems related to #2449 and #2407, perhaps others.

[Exirel]

Is this in a state I could check out on one of my bots for testing? Seems related to #2449 and #2407, perhaps others.

As said on IRC: yes go on! And btw, #2430 is approved and ready to merge, that would make this PR only 1 commit. 😄

[dgw]

Had to wait quite a while for my test bots to time out, but they both disconnected due to ping timeout at the same minute, and both reconnected automatically about 30 seconds later with no manual intervention. Win for this patch!

Both affected bots were started a few seconds apart, and pinged out a few seconds apart, pointing to an actual, brief, network issue as the root cause. In both bots' logs (example below), there is clearly time between the final PING and "Reached timeout" for another PING attempt to avert closing the socket after only one PONG is missed. It makes me think Sopel's ping timeout detection might warrant some further tweaks, perhaps in 8.1.

[2023-05-19 09:29:34,886] sopel.irc.backends   DEBUG    - Sending PING after 54.0s of inactivity.
[2023-05-19 09:30:29,037] sopel.irc.backends   DEBUG    - Sending PING after 54.0s of inactivity.
[2023-05-19 09:31:35,074] sopel.irc.backends   WARNING  - Reached timeout (120.0s); closing connection.
[2023-05-19 09:31:35,075] sopel.irc.backends   DEBUG    - Read task was cancelled.
[2023-05-19 09:31:35,076] sopel.irc.backends   DEBUG    - Shutting down writer.
[2023-05-19 09:31:40,743] sopel.irc.backends   ERROR    - Unexpected error while shutting down socket.
[2023-05-19 09:31:40,753] sopel.irc.backends   DEBUG    - All clear, exiting now.
[2023-05-19 09:31:40,758] sopel.irc.backends   INFO     - Connection backend stopped.
[2023-05-19 09:31:40,758] sopel.bot            INFO     - Shutting down
[2023-05-19 09:31:40,759] sopel.bot            INFO     - Stopping the Job Scheduler.
[2023-05-19 09:31:40,865] sopel.bot            INFO     - Job Scheduler stopped.
[2023-05-19 09:31:40,866] sopel.plugins.jobs   DEBUG    - Successfully unregistered all jobs
[2023-05-19 09:31:40,866] sopel.bot            INFO     - Calling shutdown for 2 plugins.
[2023-05-19 09:31:40,866] sopel.bot            DEBUG    - Calling sopel.modules.find.shutdown
[2023-05-19 09:31:40,867] sopel.bot            DEBUG    - Calling sopel.coretasks.shutdown
[2023-05-19 09:31:40,867] sopel.cli.run        WARNING  - Disconnected. Reconnecting in 20 seconds...
[2023-05-19 09:32:00,908] sopel.bot            INFO     - Loading plugins...

[dgw]

Base branch gymnastics to make GH update its (cached, I assume) idea of which commits are new in this PR. Can merge now.