Create (private, internally adversarial) performance fork of `cdapython`

[abradyGDIT]

• access DB directly via a DBI library or whatever they're calling it these days instead of having the API mediate queries and results
• should generate performance differences, observations of which will hopefully suggest possible targeted optimization of any trimmable API overhead