int64-timestamp-dependent test vs. --disable-integer-timestamps

Commits a7145f6bc et al. added a test to verify integer overflow
detection in interval_mul.  The buildfarm has now reminded me that
you're not going to get integer overflow if timestamps ain't integers,
cf
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=mandrill&dt=2019-11-08%2019%3A42%3A32

I think the most expedient answer is just to remove that test case
in the pre-v10 branches.  It's already served its purpose by showing
that the rest of the buildfarm is OK.  I'd work harder on this if
--disable-integer-timestamps were still a live option, but it's
hard to justify any complicated solution.

            regards, tom lane

[ wanders away wondering if we should have more than one critter testing
--disable-integer-timestamps ]