Cleanup P3A clent code / get rid of PROCHLO ?

[iefremov]

Initially, the P3A system was designed to utilise PROCHLO for additional privacy, however PROCHLO backend seems abandoned and P3A flies as is. We should consider removing PROCHLO bits from the client side to drop unused legacy and craft better pings (e.g. drop ascii string -> protobuf -> base64 conversion).

[iefremov]

cc @too4words @keur @porteron

[iefremov]

looking forward for resurrecting PROCHLO, reopen if PROCHLO is not in production in 2022 :)

[stephendonner]

@iefremov / @rillian: can you help QA by providing a test plan we can run & maybe key off of? Thanks!

[rillian]

My understanding is that this is just changing internal code paths for P3A message generation, so observable behaviour should be the same, or at least similar. Helpful QA might be:

• Verify (e.g. with a network monitor) that the same P3A message contents are sent for a particular profile. This need not be exhaustive. Spot checks and message counts without changing any preferences will validate the code changes.
• Verify that metadata tags are still stripped appropriately: On linux the browser should not report a referral code, a browser installed in a minority locale (Singapore? Nigeria? New Zealand?) should not report a region, and so on.

[LaurenWags]

Labelling QA/Blocked until discussion in https://bravesoftware.slack.com/archives/CDNJ9SVUL/p1637170701059200 is resolved

[rillian]

Unblocking and setting QA/No based on discussions. It would be helpful to have some spot testing of what the browser is sending, but data coming into the backend looks similar, so I think we can defer QA until #15967 is complete.

[LaurenWags]

Sounds good @rillian - will proceed with this unless we hear differently from @iefremov 👍🏻

[GeetaSarvadnya]

Performed a spot check on the client-side and ensured things are working as expected and nothing is regressed.