Tests linearizability with threads

[n-osborne]

Adapt lin_prop for domains and threads so that we can reuse check_seq_cons.
Tests for linearizability with thread are added in src/lin_test.ml

[jmid]

Great - thanks! 👍‍ 😀

A couple of random questions:

1. Have you tried running the tests locally to see if the tests finds thread-unsafe ones (refs, clist, ... perhaps)?
2. I can see why we need to communicate results back to the original thread, but I'm a bit hesitant about the
   new dummy : res value, as it adds another definition for a client to specify.
   I would therefore like to understand your reason for choosing this design:
   Did you observe problems with triggering errors or shrinking when the threads would allocate, e.g., compared to
   building up a result list and storing it in a shared ref when a thread is done?
3. The domain tests are repeating each test instance to "make the output more deterministic" 😉
   Did you observe whether the thread tests need less/same/more repetitions to trigger and produce small counterexamples?