Add coverage test of pull requests

[RLumSK]

Since we have already reached an excellent coverage value, we should implement a check of the coverage running against pull requests to avoid that new code is insufficiently tested. Currently, I would refrain from doing this for pushes (for convenience only).

[RLumSK]

Done and closed

[mcol]

I'm reopening as this is not working, for example it didn't catch the loss of coverage noted in #638. It seems that the logic around the .rds file that's supposed to store the coverage data is a bit messed up, as it seems that now coverage is run twice on the same branch. So the time for each coverage workflow has doubled without benefit.