Make jdaviz unit converter more robust to different data types

[astrofrog]

• Revert Temporarily pin glue-core<1.19.0 to fix test failure #2809 (since it is not upstream bug after all)
• Make sure TST: test_cubeviz_aperphot_cube_orig_flux fails with RuntimeError: NULL error object in wcslib #2805 is no longer issue in the PR to fix this.

The issue highlighted in #2805 was identified as being due to the new glue-core v1.19.0 release. In fact, there is no bug in glue v1.19.0 as such, but there are now more cases where the unit conversion machinery is called. However, the unit converter in jdaviz is not sufficiently robust, and the following fails if the data is a 2D image with no spectral component:

jdaviz/jdaviz/app.py

Lines 91 to 98 in 620ccfe

	def to_unit(self, data, cid, values, original_units, target_units):
	# Given a glue data object (data), a component ID (cid), the values
	# to convert, and the original and target units of the values, this method
	# should return the converted values. Note that original_units
	# gives the units of the values array, which might not be the same
	# as the original native units of the component in the data.
	if cid.label == "flux":
	spec = data.get_object(cls=Spectrum1D)

To work with glue-core v1.19.0, the unit converter will need to be made a little more robust and not assume that any dataset that has a flux also has a spectral axis.

There is a separate issue here that 0-d WCSes should not emit errors deep in astropy.wcs but that is another matter (astropy/astropy#16298)

🐱

[pllim]

I am not satisfied with #2812 but not sure how to proceed.