Some missing error handling in parse_cs

[embray]

In the process of investigating #2970 I ran across what I think is an unrelated buglet.

In my testing I ran

check_conesearch_sites(url_list=None)

which seems to go on to download a whole bunch of different catalog resources and look up their conesearch test queries (I'm sort of learning how this all works as I go so please correct me @pllim :) However, in the process it crashed parse_cs because at least one resource, given by the ID ivo://CDS.VizieR/J/ApJ/780/34 contains a testQuery element with only an <sr> parameter specified, and no <ra> or <dec>. Based on my reading of this, and some of the discussion here it seems the RA, DEC, and SR parameters must be provided for a valid cone search (duh), though the Resource schema itself doesn't make that entirely clear... :(

In any case I think we should:
a) Report the broken resource and
b) Fix the code to catch this kind of error (the code is currently assuming any testQuery will at least have a valid set of parameters).

[embray]

I'll add, this hasn't normally come up in testing since the test uses just a subset of sites for conesearch_master_list, and I was running the full list, I guess.

[pllim]

You generally do not want to validate all the services. I only used that option when I was doing a "first pass" to make a list of "good" services. There are a lot of invalid services in the master list (the last I check, there were over 10k total, good and bad together, mostly bad). That is why I have a short-listed list with the relatively "good" ones, and there are functions for users to build their own trusted list.

To account for all possible kinds of errors that the registry can have because the providers did not care to update or correct their info is going to open a can of worms. And I have tried contacting various providers at the start of this project. Most of them did not reply because they don't have resources to fix the problem, or the email got bounced back.

tl;dr - I'd rather mark this as "won't fix".

EDIT: The fix for this particular case is actually trivial. I'll submit a PR.

[embray]

@pllim Right, I'm not saying we need to be able to support every possible invalid service; just prevent the code from crashing in this instance.

[pllim]

@embray , yes, I agree with you now.