Avoid labeling race conditions

[timokau]

This is especially relevant after #76. When somebody leaves multiple comments in a pull request review, all of them (and the final review submission) can trigger the same -> awaiting_reviewer transition. They will race, and all but the first one will raise an exception since they try to re-remove the old label.

Not sure how best to prevent this. Just ignoring the error is an option, but not a very nice one. Another option would be to re-fetch the issue before setting a label, but that only tightens the window for the race condition.

[timokau]

Discovered in NixOS/nixpkgs#92697 (review).

[timokau]

One possible fix might be to just re-fetch the issue and retry setting the status when label-removal fails. Alternatively, just give up at that point. Both options are not perfect, but will at least avoid inconsistent labels (2 statuses at once).