Refactor the ArtifactInfo class

[AthreyVinay]

An effort towards resolving: #4055 (part-1)

following changes made:

1. All subclasses of ArtifactInfo now removed
2. Updated attributs of ArtifactInfo to what was discussed during the artifact plugin design sync (id, version, name, location, provider). The __str__ of this container represents: <provider-id>-<nvra>
3. Introduced Version object and its subclass (RpmVersion with its factory methods) to represent nvra & nevra construction
4. Refactored all classes (src and tests) where the above was referenced
5. Refactored artifacts propeorty to get_installable_artifacts method. Reason being only persist information about artifacts that would be downloaded i.e. list from providers that "download" and not "enable".
6. Removed ArtifactInfoT generic since see point 1.

couple more to finish before moving to review:
1. move ArtifactInfo to someplace else visible to package_managers too. The providers shall be responsible to create ArtifactInfo objects.

[psss]

Seems that the class is not used. Interesting that it was not caught by ruff. Ah, we have it disabled:

• Enable the ruff check for unused imports #4566

[AthreyVinay]

Shall I wait for it to be fixed in the above pr?

This import is needed to register the provider in the plugin registry. Just realised after trying to remove it.

[LecrisUT]

That is quite funky. There is a load plugin method, but I did not look into how that works. Do you have the bandwitdh to check it?

[AthreyVinay]

Maybe to unblock this - we can look at adding this to the housekeeping sprint?

[AthreyVinay]

ok... so did a bit of digging.
there's a _explore_packages() hidden in explore() that loads plugins on start... however this does not get called automagically for unit tests, we need to explicitly call it to load the providers. Added a fixture for this. See: 0119fe4
Maybe we could try lazy-loading in the plugin registry (so it auto-discovers on first access)? Obviously as a separate change though.

[LecrisUT]

Fine for now, there are some pytest constructs that can avoid adding it manually to each test (define a fixture to be used under a conftest scope), but we can get to that later.

[AthreyVinay]

'm a bit confused now - isnt that what I have pushed? a fixture _load_plugins() and making artifact_provider() depend on it. Am i missing something?

[AthreyVinay]

/packit build

[LecrisUT]

Just a few other minor comments, otherwise it's fine from my side. Could reorganize where the class definitions are, but I will leave it to you to decide when to do so.

[vaibhavdaren]

LGTM

[psss]

It's a bit late for the release and this change does not directly affect users.
Let's not block on it and merge as one of the first things in the next sprint.