Speed up semgrep on pre-commit

[gasman]

Semgrep is quite slow compared to our other pre-commit hooks, taking around 2 seconds to complete on my machine. Try to improve this a bit by:

• only running it on *.py files
• skipping the version check (saves about 300ms)

We're currently only using semgrep to enforce old-style string formatting on translation strings, which comes up infrequently enough that we might want to consider leaving this out of pre-commit and just relying on CircleCI to catch it instead. Will see how this improvement pans out first, though...

[squash-labs]

Manage this branch in Squash

Test this branch here: https://gasmanadminspeed-up-semgrep-a9ieh.squash.io

[loicteixeira]

Yeah, we use semgrep at work too, and it can be quite slow (but we do have many rules). Hopefully that helps.

If not, I agreed, this could run in the CI. The feedback loop would be a bit slower, but as you said, it's infrequent enough.