Flesh out characteristics in desk review workbooks

[jeancochrane]

This PR makes some more updates to the desk review workbooks in line with the feedback we got last week:

• Adds AC to the Card Detail sheet
• Adds AC, # Res. Units, and # Comm. Units to the Comparables sheet so its chars match the Card Detail sheet
  • Adds # Beds, Exterior Wall, Basement Type, Basement Finish, AC, Heat, and Stories to the PIN Detail sheet so its chars match the Card Detail sheet
  • Since all of these chars (except Ext. Wall and Stories) are card-level chars, we aggregate them by PIN using the card-level data and concatenate the resulting values into comma separated strings

Closes https://github.com/ccao-data/model-res-avm/issues/217.

[jeancochrane]

@dfsnow Before I merge, would you be able to spot check the tweaks in https://github.com/ccao-data/model-res-avm/pull/218/commits/21806e444ba58068c3266145fd984c7446ab13ce to make sure I'm understanding the intersection between class and APTS/NCU correctly? My understanding is:

• meta_class is set at the PIN level, while char_class is set at the card level
• We filter char_apts and char_ncu by char_class in the ingest stage for training and assessment data
• char_ncu is set to 0 when char_class != "212" in the ingest stage

Based on this I made the following changes:

• Use and display char_class instead of meta_class for cards and comps (which are defined on the card level)
• Don't filter char_apts by class, but do call format_char_apts for all 3 data sets to convert to numerals
• Continue setting char_ncu to NA for all 3 data sets when the class code is not 212 (to account for the fact that the ingest stage sets it to 0 in these cases)
  • Per the first bullet above, we're now doing this filtering on char_class instead of meta_class for the cards and comp cases

[dfsnow]

@dfsnow Before I merge, would you be able to spot check the tweaks in 21806e4 to make sure I'm understanding the intersection between class and APTS/NCU correctly? My understanding is:

• meta_class is set at the PIN level, while char_class is set at the card level
• We filter char_apts and char_ncu by char_class in the ingest stage for training and assessment data
• char_ncu is set to 0 when char_class != "212" in the ingest stage

Based on this I made the following changes:

• Use and display char_class instead of meta_class for cards and comps (which are defined on the card level)
• Don't filter char_apts by class, but do call format_char_apts for all 3 data sets to convert to numerals

• Continue setting char_ncu to NA for all 3 data sets when the class code is not 212 (to account for the fact that the ingest stage sets it to 0 in these cases)

  • Per the first bullet above, we're now doing this filtering on char_class instead of meta_class for the cards and comp cases

Yep, that perfectly tracks with my understanding. Let's merge it and get the new sheets!