Test suites for the PR labels?

[Sheshuk]

We have many tests now, which are being run on every PR.
While this is good for staying consistent, it makes a lot of unnecessary testing occuring: when some small correction is applied to documentation, we still check that all of our models are downloadable etc.
Probably there is a way to separate our test suites (like we already do for snowglobes, assigning pytest.mark in the test files) into separate sets, which would be run on the PR only when it has a certain label.
Something like that:

• default tests for all PRs (some basic checks without many computations)
• ModelRegistry - test that models can be downloaded
• SNOwGLoBES - test rate calculation for some of the models (no need to download or use all of them)
• SupernovaModel - test that all models can be instantiated etc.
• documentation - try to build the documentation and probably provide it as an artifact of the PR
• etc

Also probably there is a way to define the tests suites based on the files, modified in the PR.

This can reduce the usage of our runners, and address #238

Maybe you can add more ideas