Add property-based tests

[barrucadu]

There are certainly some things which could benefit from property testing, if someone was willing to write the Arbitrary or Listable instances. Here are a few I spotted skimming through Test.DejaFu.SCT.Internal:

• \ds tid act -> canInterrupt ds tid act ==> canInterruptL ds tid (rewind act)
• dependentActions is commutative
• \mem ds tid1 ta1 tid2 ta2 -> dependent mem ds tid1 ta1 tid2 ta2 ==> dependent' mem ds tid1 ta1 tid2 (rewind ta2) (I think that should be the case...)

I suspect the various Test.DejaFu.Conc.Internal modules and Test.DejaFu.Common will be fruitful as well. Depending on how crazy you want to get, you could even check that Eq and Ord instances are consistent, where one or both is hand-written.

This is worthwhile because there's a lot of logic which differs only in whether it accepts a ThreadAction or a Lookahead value. The ThreadAction case is supposed to be the more accurate case, whereas the Lookahead case is supposed to be the more pessimistic case, so checking that (eg with implication if it's a predicate) would help make sure there aren't any mistakes.

[barrucadu]

Added some tests, but it would be nice to have some for the memory implementation as well. However, I'm not really sure how to best add those, as leancheck doesn't do monadic properties.

At the moment runST would work, but in implementing bound threads, everything is changing over to IO. Maybe that module should keep using MonadRef, as it won't acquire any IO-specific code.

[barrucadu]

An alternative to fitting MonadRef back into the bound branch would be to define an orphan MonadConc (ST t) instance for testing, which simply defines CRef, readCRef, and atomicWriteCRef, leaving everything else undefined.

This is probably a far smaller change.