update industry_spec_key() so it is not based on sector string length

[catherinebirney]

• Update how target_naics are created so it is not based on the sector string length, instead determine target_naics using df of sectors by sector_length
• addresses error mapping to special sectors (e.g. F010) for summary models in sector_aggregation
• also reformat industry_spec dictionary so key is target naics length and sectors/values are in a list

[catherinebirney]

@matthewlchambers in my update of mapping source_naics to target_naics I did include the option of including non_naics (e.g., "non_naics: ACC") but I didn't quite understand what was happening originally so I'm not sure I included it correctly and I'd appreciate if you would check that the non_naics map correctly

[matthewlchambers]

@matthewlchambers in my update of mapping source_naics to target_naics I did include the option of including non_naics (e.g., "non_naics: ACC") but I didn't quite understand what was happening originally so I'm not sure I included it correctly and I'd appreciate if you would check that the non_naics map correctly

Yep, I'll check that! I like this, it's much smoother than what I had before!

[matthewlchambers]

I really love the new crosswalk, and the use of melt! This is sooo much easier to follow than what I had before, in addition to being more robust.

I just pushed 3 commits to this branch, making the following minor changes:

1. I changed the code that loops through industry_spec and updates target_naics to make the logic a little simpler.
2. I updated the code for non_naics to make it clearer than my original version and (incidentally) allow the user to specify a list of non-NAICS codes if needed.

[catherinebirney]

Looks good, thanks @bl-young and @matthewlchambers. I will merge.