Suggestion: Do not subtract from approvals if there were no code changes

[lorenzleutgeb]

When someone very clearly expresses their approval in a comment, but does not use the GitHub feature to signal approval, I sometimes increase the label-based approval counter myself (to attract attention of potential mergers). Of course I could ask the person that commented favorably but did not use the GitHub feature to do exactly that, but this requires another roundtrip (I comment, they receive a notification on their end, they approve, and only then future runs will increase the label-based counter, or I do it myself).

I noticed that in these cases, the tool sometimes decreases the number of approvals, see e.g. NixOS/nixpkgs#311986 where there was approval clearly expressed in a comment, I set the label, and later @wegank ran the tool which decreased the counter.

I therefore propose that under the condition that there were no new commits in the PR, the number of approvals should not be decreased automatically.

If people that increase the label-based approval counter manually (like I do) are honest, this should work out pretty nicely.

[wegank]

I therefore propose that under the condition that there were no new commits in the PR, the number of approvals should not be decreased automatically.

The counter should at least decrease if a user dismissed their previous approval, though.

I don't have a good idea yet how to implement the feature rigorously without tracking the timeline, which costs an extra REST API request per approved PR...