[MISC] Miscellaneous cleanups

[astrojuanlu]

Issue

Ended up accumulating lots of cleanups in #87, so I decided to send a separate pull request:

Note: This does not include adapting to the new backup helpers, because it turns out those need a writable temporary directory. Instead of introducing some interim hack, I prefer to do both adaptation + separation of storage in a single PR.

Main changes:

• Fix backup methods in VM and K8s child classes so that they have a compatible signature with the parent class
• Fix type hints of TLS handling code (see also [MISC] Fix tls_ca_chain type object-storage-integrator#170)
• Visually separate TLS handling in backups code
  • This introduces some repeated error handling, with the benefit of improved readability
• Make Ceph backup test amenable for local development
  • feat: Use s3-integrator from track 2/ with the newer object-storage lib #88 had introduced some changes that were not properly accounted for in this test
• Improved support for ty
  • Including some ty configurations in pyproject.toml but also some actual typing improvements
• Minor improvements in observability, metadata, typos

Solution

Checklist

• I have added or updated any relevant documentation.
• I have cleaned any remaining cloud resources from my accounts.

[astrojuanlu]

Upgrade tests failing with

unit-mysql-0: 21:20:34 INFO unit.mysql/0.juju-log charm_refresh:Refresh incompatible because new version of MySQL ('8.4') and/or charm (CharmVersion('8.4/1.22.0.post13.dev0+6d221716')) is not compatible with previous version of MySQL ('8.4') and/or charm (CharmVersion('8.4/1.15.0'))

The rest seem just flaky or slow, I'll restart them a few more times.

[paulomach]

Nothing polemic, better optic

[sinclert-canonical]

I see we are not fully decoupling the generation of the CA chain location from the creation of the backup temporal directory, as discussed online. I assume there is a good reason for this, because last time we spoke we were on the same page about splitting them.

The proposed approach is alright to me.

---

The same comments left for the K8s implementation, also apply to the VM one.

[astrojuanlu]

I see we are not fully decoupling the generation of the CA chain location from the creation of the backup temporal directory, as discussed online. I assume there is a good reason for this, because last time we spoke we were on the same page about splitting them.

Yes. On one hand, the CA chain file creation is already split:

mysql-operators/machines/lib/charms/mysql/v0/mysql.py

Line 2451 in e10ede0

def create_ca_pem_file(

And on the other hand, I was misguided - I thought the CA chain file creation was the exception, but it wasn't: executing the wrapper as root was. More rationale in the description of #257, happy to elaborate more or explain synchronously if needed.

[astrojuanlu]

Tests were passing before, so I'm merging and rebasing #257.