Update CI script

Description

Following time-out problems in CI jobs, we tried to force different runner tags:

• saas-linux-small-amd64 (default): OK
• saas-linux-medium-amd64: OK
• No tag (current version - equivalent to default): OK

The root cause of the original problem remains unknown. In the meantime, the CI script has been slightly refactored.

References

• https://gitlab.com/cosapp/cosapp/-/jobs/11633830688
• https://gitlab.com/cosapp/cosapp/-/jobs/11642225510

Note: the jobs always stalled on a particular test, test_LinearDoE_multiprocessing[WorkerStartMethod.SPAWN-4], so it was originally assumed that the small Linux runner (2 CPUs, 8GB of RAM, from https://docs.gitlab.com/ci/runners/hosted_runners/linux) was not able to handle 4 processes. It turns out it can.