Consensus and documentation on testing methods

[JamesGuthrie]

We have too many options for testing, and it's not clear which is best suited to what:

• pgx tests
• end-to-end golden tests
• end-to-end SQL tests
• end-to-end Rust tests

We should understand what the tradeoffs between all of the above testing methods are, potentially reducing or combining the different methods (note: if we make it possible to run e2e without docker #276, this may increase the amount of stuff that we can combine). Finally, we should have documentation which recommends which methods should be used for what.