BUG: fix jplspec build lookup

[bsipocz]

This should fix the devdeps CI job.

Given the columns have a well defined name, we should just select columns based on names. Note, this hasn't caused any issues up until the 6.1dev cycle. I haven't done any bisecting to pinpoint what made it fail upstream as this fix will work for us here.

ps: no need to add a test as currently existing tests alerted us about this regression

[pllim]

I feel like there is some unnecessary steps from Table to list/dict back to some other table, but I guess fixing that is out of scope. If this change is enough to make the code be compatible with astropy 6.1.dev, then LGTM.

Though I am curious to see if @taldcroft knows why [:] now breaks for you in astropy 6.1.dev but not in 6.0.

[bsipocz]

Adam has some work in progress for fixing how the lookup dictionary is used, etc, so I would not dive into anywhere beyond fixing the obvious bug.

I also don't think I have the full understanding of the desired parsing and search capabilities for these lookups and whether a dictionary subclass or a Table subclass, or not subclassing at all is the right way to do it.

[keflavich]

This change makes sense to me. I think it's more likely they'll change the order/number of heading parameters than it is that they'll change the NAME or TAG headings. I believe the original motivation was that I had the opposite expectation (i.e., I thought it more likely they'd change the capitalization, say, than the order) - and probably there was no reason to expect that.