Add invariant checking

[sjakobi]

Addresses #366.

---

TODO:

• Add more tests
• Document invariants
• Extract validation code into D.HM.I.Debug?!

[sjakobi]

I've checked that this tooling could have caught #420. Sample output:

        intersection produces valid HashMap: FAIL
          *** Failed! Falsified (after 61 tests and 10 shrinks):
          fromList [(K {hash = 265, _x = D},()),(K {hash = 16649, _x = A},())]
          fromList [(K {hash = 265, _x = D},()),(K {hash = 16649, _x = A},())]
          Invalid (INV2_misplaced_hash 265) (Cons 0 Root) /= Valid
          Use --quickcheck-replay=665566 to reproduce.
          Use -p '/intersection/&&/Data.HashMap.Lazy.difference and intersection.intersection produces valid HashMap/' to rerun this test only.

[sjakobi]

Hmm, is there another bug in the new intersection implementation?! oO

[sjakobi]

Hmm, is there another bug in the new intersection implementation?! oO

Probably not. I'm getting the same errors with the naive old intersection implementation. I must have made a mistake in the validation logic.

[sjakobi]

This should be ready for review. I'll hold off on adding more validity tests until I've refactored the existing property tests a bit.

[treeowl]

Greater than, or just equal to?

[sjakobi]

l is 65 when the deepest leaves are checked on a 64-bit computer.

[treeowl]

Huzzah!