Optimise & refactor the testing pipeline

[Gedochao]

Version(s)
v1.3.2 (although I assure you it'll be relevant for a long while)

General context
Our testing jobs have gotten awful long & flaky, while we keep on adding more tests. The time has come for us to deal with this, as false-positive failures on the CI are now a daily occurence, I've begun to expect a restart or 2 on every PR or release, and this isn't going anywhere good.

Describe what needs to be done and why
Here's what I'd suggest:

• add the --offline flag in all integration tests which don't need to check online activity, to eliminate unnecessary checking of dependencies and other slowdowns/timeouts due to connection quality on the CI.
• verify if some tests haven't gone deprecated and could potentially be dropped/updated/optimised.
• eliminate inevitable duplicate tests.
• go through past test reports, identify the flakiest of tests and try to make them pass consistently.
• (optional, as this a tall order) move some of the integration tests to unit tests, where and if applicable
• The generate-windows-launcher job fails intermittently on cs download timeout #2930
• Migrate from the custom M1 runner to the official Github one #2936
• (after Migrate from the custom M1 runner to the official Github one #2936 is done) split the test suites we run on M1 similar to other platforms and run them for all workflows, rather than just on the main branch.