Updates for CODI 2.0

[dehall]

This PR introduces changes for CODI 2.0:

• Updates the data model extraction to support either the old data model (ie, the codi.IDENTIFIER table) or the new data model (cdm.PRIVATE_DEMOGRAPHIC and cdm.PRIVATE_ADDRESS_HISTORY). Currently this is selected based on a command line arg, we may want to be nice and try to infer which version to use instead.
• Allows data owners that maintain their own household relationships to provide that, instead of having the household script try to infer them out of the pii.csv file. The inference process is still there when needed.
• Removes all "parent" identifiers. These are really only relevant for a pediatric population as in CODI 1.0, but in practice they weren't used anyway as only the "parent contact email" field was ever used in a schema, and that schema was disabled following QA.
• Adds PATIDs aka record_ids to the households_pii.csv format. This file will either be inferred by households.py, or created by the data owner if they maintain their own household relationships. Adding PATIDs to this makes it easier to link back to them after receiving HOUSEHOLD IDs from the linkage agent. And hopefully makes it easier for the data owner to create vs multiple correlated files.

Still TODO:

• More testing

[dehall]

I had an epiphany last night - why not just include the PATIDs column in the household_pii file? (i.e,. the list of PATIDs associated with the household) That makes it easier to crossreference HOUSEHOLDIDs to PATIDs since there are fweer files, and it's probably easier for a data owner to create too. Will update that next