Test.check_exn and more unit tests

[jmid]

When writing the shrinking logs I noticed that the Test.check_exn was wrong:
It said it would throw a Test_error exception on both failures and errors.
I figured it would be nice to add a couple of unit-tests to ensure this for QCheck2.

This PR does both of these.

As a bonus, it adds a few unit tests for QCheck, including some for the int* shrinkers
to bring the unit test status closer to QCheck2.

After having realized that the shrinker's output depend on whether the tested property holds or not,
I split them up into a successful shrink trace and a failure shrink trace,
I then noticed that our current int* shrinkers produce the same number twice at the end, e.g.,
for Shrink.int 100 tested on, e.g., (fun i -> i<100) it tries: [50; 75; 88; 94; 97; 99; 99] 🤔
This is confirmed by shrink_algo_logs/int_smaller_209609_qcheck.expected which tests 209608 twice at the end.

I'll investigate. I think that could make for another potential Shrink.int improvement on top of PR #173.

[jmid]

BTW: This PR builds on top of #172 - which renamed the QCheck2 unit test file.

[jmid]

Quick update on the int shrinking:

qcheck/src/core/QCheck.ml

Lines 470 to 477 in 189a025

	(* balanced shrinker for integers (non-exhaustive) *)
	let int x yield =
	let y = ref x in
	(* try some divisors *)
	while !y < -2 || !y >2 do y := !y / 2; yield (x - !y); done; (* fast path *)
	if x>0 then yield (x-1);
	if x<0 then yield (x+1);
	()

When x is 3, say, we will enter the loop, setting y := 1 and yielding 2.
After resuming we will then exit the loop only to yield 3-1 = 2 again.

PR #173 doesn't seem to have this problem.

[jmid]

Superseeded by #233 and #234