Increase reproducability of Lin Thread tests

[jmid]

This splits the negative tests of Lin into

• Domain tests and
• Thread tests

In order to increase reproducability it then

• uses an allocating, non-tail-recursive interp_thread function
• delays the first thread until the second is up and running
• increases the rep_count

For the latter I spent a bit of time developing a statistical test to decide whether a candidate change would
constitute an improvement. I then recorded for 10.000 arbitrary cmd-triples whether the property would hold,
and used these numbers for a z-statistic test following https://www.itl.nist.gov/div898/handbook/prc/section3/prc33.htm
The three above changes proved to yield statistically significant increases.

Things that didn't pan out

• Thread.delay instead of Thread.yield to maximize concurrency issues
• extra Thread.yields in the Thread.create lambdas
• ...