Move some Vendor-specific group attributes to data variables

[leewujung]

This PR moves some Vendor-specific group attributes to data variables to make it more efficient to lazy-load and combine echodata objects, and potentially on the writing speed to cloud. This addresses #1055 and potentially #640 (AZFP) and #1045 (though the goal changed, EK80).

The changes include:

• EK80: move xml_config which is the entire XML string from an attribute to a data variable
• AZFP: move all Vendor-specific group attributes to data variables
  • some of these variables come from the XML associated with the 01A file
  • some of these variables come from the unpacked ping by ping data
  • a common feature of these variables is that they are not time-varying (which is probably why we saved them as attributes in the first place)

I've tested combine_echodata with the new format and these are combined correctly -- that all these variables remain data variables without a dimension (constants). The combine would error out if any of these variables change across the echodata objects to be combined, which is the desired behavior since those echodata objects should not be combined.

[emiliom]

Looks good! Things are so nice and simple when you don't have to worry about having variable attributes 😆 . I like how you organized the variable assignments into blocks with a helpful but brief comment line.

Just one question: I assume that these changes have no downstream effect on calibration functions, since there are no changes to that code?

[leewujung]

Just one question: I assume that these changes have no downstream effect on calibration functions, since there are no changes to that code?

Oh right, this completely slipped my mind. The ones that I can think of on top of my head are the conversion for temperature and tilt_x/tilt_y, from the raw count combining with the coefficients in the variables (_compute_tilt, _compute_temperature). I looked them up and luckily it actually uses the parsed data dict instead of the echodata vendor-specific group entries for that, since the conversion is in the parser. So I guess the changes are not used anywhere. I'll take another look on the ones I moved, just to be sure.

[leewujung]

Ok, confirmed -- none of the attributes that I moved to be data variables are used in later calculations. :)