Improve tests and coverage

[xoofx]

Many parts of dotnet-releaser rely on running on GitHub to trigger specific actions (e.g publish...etc.). We can't easily replicate this in a sandbox test. There is some partial mocking, but that wouldn't test these parts.

Also, the tests are running the command dotnet-releaser as a separate executable, so the coverage doesn't capture these runs.

At least, dotnet-releaser is dogfooding itself when running on the main branch, so it can capture issues, but it still doesn't help with classic tests or coverage.