Regression test for test duration

[goatgoose]

Problem:

We sometimes introduce changes/new tests which causes test runtimes to dramatically increase (see #5558). Ideally we would figure out a better way to write the test, or fix whatever issue is causing the test to take a long time to run before merging the change. However, we currently have no way to detect this at PR time.

Solution:

We should keep track of how long each test takes to run (e.g. s2n_security_policies_test.c), and add a job to CI that fails if the test duration increases too much.