feat: support coverage>=5.0

[TheKevJames]

I can't say I'm excited about this code, but at the very least this should be fully backwards compatible with how coverage<5 + coveralls work together. There's definitely going to need to be some future work to clean this up, but hopefully this should be enough to get folks unblocked from using coverage>=5.0.

[TheKevJames]

@nedbat looks like I'm getting some errors about mixing coverage versions' .coverage files (eg. see this CI run). Theoretically, this issue only affects me since I'm explicitly matrix testing coverage versions, but... any thoughts on ways I might be able to merge those results?

[TheKevJames]

@nedbat is there anything I missed in the public coverage v5+ API which would support this? Specifically, its the following block that I couldn't work out how to do without private APIs:

for l1, l2 in analysis.arcs_executed():
    if l1 in branch_lines:
        # SNIP

[nedbat]

@TheKevJames let's talk about this in another place. Clearly the Coverage public API doesn't give you everything you need yet. You called out this spot, but looking at the rest of the file, there are many things you are using from coverage.py that I didn't anticipate you would need.

I've added an issue for coverage.py, let's discuss the details there: coveragepy/coveragepy#921