BUG: fix parsing by filtering out junk lines

[bsipocz]

This closes #2717

I'm on the fence whether this needs a changelog or not, I suppose it does as lines from the docs stopped working due to these junk lines, e.g. this one

>>> response['FREQ'].quantity

[bsipocz]

This also fixes all the doctesting failures we currently see in the docs (some are due to these lines, but some are due to upstream changes)

[codecov]

Codecov Report

Merging #2719 (66f2099) into main (1ba336e) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main    #2719   +/-   ##
=======================================
  Coverage   65.74%   65.74%           
=======================================
  Files         233      233           
  Lines       17844    17844           
=======================================
  Hits        11732    11732           
  Misses       6112     6112           

Impacted Files	Coverage Δ
astroquery/jplspec/core.py	91.78% <ø> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[keflavich]

except for minor typo, looks great

[mkelley]

This looks good to me, except that the LGINT units comparison failed for me locally:

Package versions: 
Numpy: 1.21.4
Scipy: 1.7.1
Matplotlib: 3.5.2
h5py: 3.4.0
Pandas: 1.3.3
Astropy: 5.0.1
regions: not available
pyVO: not available
mocpy: not available
astropy-healpix: 0.6
vamdclib: not available
astropy-helpers: 4.0.1

Using Astropy options: remote_data: any.

rootdir: /home/msk/Projects/astroquery, configfile: setup.cfg
plugins: cov-2.12.1, typeguard-2.13.3, xdist-2.5.0, arraydiff-0.3, requests-mock-1.9.3, remotedata-0.3.2, hypothesis-6.21.6, anyio-3.6.1, forked-1.4.0, mock-3.6.1, openfiles-0.5.0, filter-subpackage-0.1.1, astropy-header-0.1.2, asdf-2.8.1, doctestplus-0.11.0
collected 1 item                                                                                                                                                                         

docs/jplspec/jplspec.rst F                                                                                                                                                         [100%]

======================================================================================== FAILURES ========================================================================================
_________________________________________________________________________________ [doctest] jplspec.rst __________________________________________________________________________________
027 .. doctest-remote-data::
028 
029    >>> from astroquery.jplspec import JPLSpec
030    >>> import astropy.units as u
031    >>> response = JPLSpec.query_lines(min_frequency=100 * u.GHz,
032    ...                                max_frequency=1000 * u.GHz,
033    ...                                min_strength=-500,
034    ...                                molecule="28001 CO",
035    ...                                get_query_payload=False)
036    >>> print(response)
Differences (unified diff with -expected +actual):
    @@ -1,4 +1,4 @@
    -    FREQ     ERR    LGINT   DR   ELO   GUP  TAG   QNFMT QN' QN"
    -    MHz      MHz   nm2 MHz      1 / cm
    +    FREQ     ERR    LGINT   DR   ELO    GUP  TAG   QNFMT QN' QN"
    +    MHz      MHz   MHz nm2      1 / cm                          
     ----------- ------ ------- --- -------- --- ------ ----- --- ---
     115271.2018 0.0005 -5.0105   2      0.0   3 -28001   101   1   0

[keflavich]

I think the lgint comparison is related to astropy version - possibly this PR:
https://github.com/astropy/astropy/issues?q=order++units

[bsipocz]

Oh, great. Then I need to double check whether I use the same version as ci.

[bsipocz]

OK, so this is the PR that brake stuff, and the suggested workaround for a global pin in setup.cfg won't do it for us:
astropy/astropy#14439

We do run remote-data tests for the latest astropy release (5.2.x currently), and with the dev version, so I would vote for making these changes for the dev test and live with the breaks for the windows online tests until 5.3 comes out rather than adding hacks.

This change affects the jplspec module, and cdms, and svo_fps, and only the narrative documentation. Fix for the new formatting is in this PR and in #2721.

[bsipocz]

Side note: I'm very much looking forward to use scientific-python/pytest-doctestplus#197, so our rst doctesting won't be bogged down at a doctest failure for a print command.