Fix test failure from updated MUPS model

[taldcroft]

Description

This fixes the failure below, which is due to a change in the number of "bad" samples after the underlying MUPS thermal model was changed in chandra_models master since the test was written. The original test was badly designed and susceptible to this failure.

_______________________________ test_mups_valve ________________________________

    def test_mups_valve():
        colnames = ['vals', 'times', 'bads', 'vals_raw',
                    'vals_nan', 'vals_corr', 'vals_model', 'source']
    
        # Use the chandra_models 6854df4d commit for testing. This is a commit of
        # chandra_models that has the epoch dates changes to fully-qualified values
        # like 2017:123:12:00:00 (instead of 2017:123). This allows these regression
        # tests to pass with Chandra.Time 3.x or 4.0+.
        dat = fetch_eng.MSID('PM2THV1T_clean_6854df4d', '2020:001:12:00:00', '2020:010:12:00:00')
        assert dat.unit == 'DEGF'
        assert len(dat.vals) == 36661
...
        # Check using default master branch
        dat = fetch_eng.Msid('pm1thv2t_clean', '2020:001:12:00:00', '2020:010:12:00:00')
>       assert len(dat.vals) == 36240  # Some bad values
E       assert 36639 == 36240
E         +36639
E         -36240

Interface impacts

None

Testing

Unit tests

• Mac

Functional tests

No functional testing.

[javierggt]

I ran the tests manually with the current release and with this branch and can confirm this PR fixes it.