Credo should run once per commit instead of once per file

[MikeBenza]

Credo takes about 2 seconds to run per file. You can pass more files to credo but the output then needs to be parsed.

[MikeBenza]

Running credo once on a set of files would also require making it so that the temporary old version of the file is in the correct directory. Supposing you changed a/foo.ex and b/foo.ex, they'd need to be in their correct directories otherwise one would overwrite the other.

[StandardMikeBenza]

...furthermore, parsing would require the output to be JSON and there's no guarantee that jq is present. It'd be annoying to have a version that works fast with jq and slow without.