upstream ci: Use a single random seed for spliting tests

Dependind on how long it took for the jobs to start, a different seed
would be used to group tests and tests could either repeat or not be
selected at all.

By using a seed based on the day the test run reduces the chance of
using different random seeds, and still allow for the tests to be
executed in a different order.

The execution in different order is important to identify tests that
work or fail only if executed after other tests.