Separate CI job to run longer CW tests

The following discussion from !1904 (merged) should be addressed:

• @karl-wette started a discussion: (+1 comment)

  @matthew-pitkin @david-keitel I'm seeing the test_flat_evidence.py test occasionally fail when the value it tests is over the threshold of 2 sigma. For now I've now increased this threshold to 4 (default proposal) or 5 (walk proposal; currently not run). Admittedly this many not be a strong enough threshold to test for breakages, but hopefully it may still catch egregious bugs. (There's also the test_flat_prior.py test which exercises the known pulsar code.)

  Alternatively we could set up a separate CI job to run longer CW tests, similar to the existing lalinference_testjob. That might be useful for the known pulsar code, and perhaps other codes where a longer-running test is needed.