The semver checks may take a long time (2 min * 200 crates for some devices, 30 s * 200 crates for faster ones). This means publishing the crate can take hours to run and if there is a problem at a later step, we have to re-run all of the checks even if there is no code change.

We should add a mode to skip the checks.

We should also consider adding a way to resume the checks. This would look like:

1. Check that repo is in the same state
2. Skip checks for any crates we have already recorded success for
3. Check crates and record successful result
4. Maybe record errors and return those on re-run instead of re-running the cargo semver check command,