Improve error reporting for clock test

[kennykerr]

This test fails intermittently but it's hard to tell what's wrong since the error report doesn't include the failing values. This just ensures that we can see the values when it fails and perhaps gain some more insight in the failure.

[alvinhochun]

https://github.com/microsoft/cppwinrt/actions/runs/3719133191/jobs/6307790757#step:8:85

 4: -------------------------------------------------------------------------------
4: clock, time_t
4: -------------------------------------------------------------------------------
4: D:\a\cppwinrt\cppwinrt\test\old_tests\UnitTests\clock.cpp:67
4: ...............................................................................
4: 
4: D:\a\cppwinrt\cppwinrt\test\old_tests\UnitTests\clock.cpp:80: FAILED:
4:   REQUIRE( diff < 1000 )
4: with expansion:
4:   1001 (0x3e9) < 1000 (0x3e8)

Still have no idea why this fails randomly...

[alvinhochun]

Ah, I have a suspicion: time(nullptr) has one-second resolution. Even when it is called at the 999th millisecond, the value it returns should be rounded down. If it was called first, it is possible that, when calling clock::now() the wall clock has ticked to the next second. This way, the difference between the two timestamp can slightly exceed 1000 milliseconds when unlucky.

The diff is calculated by abs(clock::now() - clock::from_time_t(time(nullptr))) in the same statement. I think because the order of evaluation is unspecified, the scenario described above may happen.