Avoid divide-by-zero in munit_rand_state_at_most when max is zero

[Syeberman]

The following expression will trigger a divide-by-zero error when max is zero:

munit/munit.c

Line 971 in fbbdf14

const munit_uint32_t min = (~max + 1U) % max;

This can occur in two places. The first is munit_rand_int_range, when min equals max:

munit/munit.c

Lines 1001 to 1009 in fbbdf14

	munit_uint64_t range = (munit_uint64_t) max - (munit_uint64_t) min;
	if (min > max)
	return munit_rand_int_range(max, min);
	if (range > (~((munit_uint32_t) 0U)))
	range = (~((munit_uint32_t) 0U));
	return min + munit_rand_at_most(0, (munit_uint32_t) range);

The second is in munit_test_runner_run_test, when a parameter has a single value (possible equals 1).

munit/munit.c

Line 1617 in fbbdf14

pidx = munit_rand_at_most(munit_str_hash(test_name), possible - 1);

Put checks in place to avoid calling munit_rand_state_at_most with a max of zero.

This is the same issue referenced in #30 and #59.