This should fix the test and make the test failures (if any)
more informative.
The issue was that it was trying to extract a 46-bit segment
(the timestamp). But because the whole thing wasn't 0-padded first,
the timestamp might be shorter. So it would extract part of the next
segment too. That meant it was testing a 46-bit segment with no
guaranteed properties, with arbitrary results.
Bug: T131549
Change-Id: I11c9e86951cc1d6b186710cda6e42ee58c920db8
aef78d13d4