Ensure all desired attributes are returned by the data translators

[rosteen]

This is intended to be a round-up/continuation of a few separate discussions:

• In PR17, @eteq and @astrofrog brought up the need to have predefined masks successfully round-trip through Glue and back - currently the mask returned by the data translators is based (I think) solely on any subset regions that are defined.

• @ibusko brought up the need to have uncertainty (and another attribute that I don't remember off the top of my head) successfully round-trip.

• @havok2063 and @nmearl and I were discussing handling more complex integer masks used to denote various problems a la SDSS (perhaps via a flags attribute rather than overloading mask, which is currently intended to be boolean).

Ensuring that all desired attributes are successfully ingested into a Glue data object and returned by the glue-astronomy translators may require work outside glue-astronomy, but I created this issue here since the discussion was initially motivated by changes needed in the data translators.

[ibusko]

Masks should also round-trip successfully, besides the uncertainty.

[rosteen]

Ah, I wasn't aware of PR 19 when I made this issue - it looks like that PR implements part of what is needed here (getting masks and uncertainty from a Spectrum1D object into the Glue data object).

[nmearl]

Mask cannot round-trip currently, because they are overridden by the subset masks created in glue.