Add performance tests based on cpython build runtime

[tristanlatr]

Since #576, the system test of the standard library is now taking 47mn instead of about 35.

It should not have influenced the efficiency at this point. So I’ll consider reverting this PR.

[tristanlatr]

Ok, I've checked the action runs history and it looks like the runtime is not very stable...

It takes from 35 minutes to 1 hour. I don't know exactly what the best way forward is, I know a Github job step can run up to 6 hours. So we might want to run the cpython API docs 3 times, calculate the mean, and report if the mean is more than one hour.

Also let's note that we are not checking out a specific version of the cpython, so it's not the best for consistency.

The extension system has probably nothing to do with this.

[tristanlatr]

The primer includes a time based comparison, so I’ll close this issue.