A lot of tools are packages there and saves a bit of a hassle.

Only on schedule or if the templates are changed.

Merge the ci-templates and ci-images repositories.

Can't mix and match or override rules and when.

There's really no point in it. I usually keep the master branch
buildable and working so running pre-commit is a waste of time and a
source for failed builds that shouldn't be.

Test all of the files in the repo, otherwise most (all?) hooks will not
even run.

See 0a1e7b83.

First of all, I'm replacing the hadolint hook that I maintained with the
one from https://github.com/AleksaC/hadolint-py and the docker-compose
hook with the new one I maintain at
https://git.shore.co.il/nimrod/pre-commit-hooks. Both don't require
Docker at all (just regular Python). Also, with
https://github.com/pre-commit/pre-commit/issues/1387 and then
https://github.com/pre-commit/pre-commit/issues/1918 resolved, the
original hooks that required Docker should work anyway.

I'm keeping the gitlab-ci hook disable for now, the 401 issue is still
present and there's little point in validating the CI config while it's
running (it is running after all).

The linter changed the usage and doesn't require the --file flag
anymore.

Bats now supports junit's XML output so there's no need for it.

Solves the issue with twine complaining about syntax errors in the
long_description.

It's the supported option nowadays and I put it in the after_script
section so it won't break anything if it fails.

For including additional Bats libraries.

Just the image, cache and variables.

At this point, it's already processed by GitLab CI so there's little
point in running the same thing again.

The pre-commit ci-image doesn't have the Terraform CLI, but it can be
downloaded in the before_script part so it's not an issue.

This reverts commit 35a24a67. There's an
issue with pre-commit now, the way it gets the container ID is
incorrect. I'll open an issue and file a PR.

The issue is reported to have been solved.