Ignore BinTableHDU's and HDU's lacking shape

[imbasimba]

I tested a barrage of different FITS files and most failures were caused by BinTableHDU's or HDU's lacking shape

[imbasimba]

Also this check causes problems for the XMM fits I tested

toasty/toasty/collection.py

Lines 97 to 101 in a64a727

	keep_axes = [t.get('coordinate_type') == 'celestial' for t in full_wcs.get_axis_types()[::-1]]
	for keep, axlen in zip(keep_axes, shape):
	if not keep and axlen != 1:
	raise Exception(f'cannot process input `{fits_path}`: found a non-celestial axis with size other than 1')

Here is a RGB FITS that is failing because of the above check. I'm not sure which solution would handle this kind of FITS the best.
https://nxsa.esac.esa.int/nxsa-sl/servlet/data-action?RETRIEVAL_TYPE=PRODUCT&OBS_IMAGE_TYPE=OBS_EPIC&PROTOCOL=HTTP&OBSERVATION_ID=0044350101

[pkgw]

Hm. Well, the current check is behaving as I intended: this RGB image essentially consists of three different images (R plane, G plane, B plane), and it's defensible to refuse to process the data if the tiling isn't well-specified, i.e. the user hasn't specified which plane they want you to process. Of course, we don't provide a mechanism to specify which plane to prcess, which makes things a bit inconvenient :-)

But in these circumstances it would also be defensible to go ahead using the 0'th image (the R plane in this case), if we gave the user a big prominent warning that that was happening. And from a user experience standpoint, people generally prefer "proceed as best you can" instead of "refuse to do anything less than perfect" ...