Develop more policy around payload_bytes data toxicity

[jklukas]

We need to more formally consider the implications of "chemspill" type incidents where we ingest data that contains sensitive info we don't intend to handle. We may need to define several levels of toxicity and differing procedures as to how thoroughly we need to clean.

In particular, the payload_bytes_raw, payload_bytes_error, and payload_bytes_decoded tables will all have to be modified if analytics users have access to them. We have been trying to make those tables more accessible to users for debugging, but it may not be worth it if it means we're on the hook for more aggressive purging in the case of incidents.