Allow ignoring individual lines for coverage analysis

[CyberShadow]

This adds a work-around to the non-deterministic coverage fluctuations
in std.parallelism. Lines with nocoverage on them will be excluded
from coverage analysis, and be considered as not containing any code.

[JackStouffer]

Seems like treating the symptom and not the cause. Why not fix the tests?

[wilzbach]

Seems like treating the symptom and not the cause. Why not fix the tests?

Well, feel free to have a go at it (warning: it's not as easy as it looks!)!
Of course if you can provide a test that always covers this line this would be preferred ...

[CyberShadow]

I've been told people have tried and didn't see an obvious simple fix, but feel free to have a go at it.

Either way, there will inevitably be non-deterministic conditions and error conditions that are too difficult to simulate, so it's good to have a mechanism to handle that in a semi-reasonable way without breaking the CI status of half of all PRs randomly.

[wilzbach]

As mentioned imho this is a very valuable addition as it avoids a CodeCov failure for a transient line(s) and thus the overall red cross for PRs, which is very confusing and frustrating for contributors.

[wilzbach]

I am inclined to merge this as (1) it's a very simple solution for a long outstanding problem about which many contributors have complained as it's very hard to figure out that the PR is displayed has "failed" due to a unrelated line in std.parallelism being only sometimes hit. Moreover, (2) except for @JackStouffer no one seems to care much about this addition, so please if someone has a better solution in mind that he would like to do instead, please raise your voice, otherwise imho we should should go with this in the next days ;-)