Review usage of crate::receive::Error

[spacebear21]

Follow-up from #277 (review)

Server errors are meant to be returned as a response, but are being returned in places that don't always make sense, e.g.:

rust-payjoin/payjoin/src/receive/mod.rs

Line 486 in 4bc8fe3

return Err(Error::Server("Output substitution is disabled.".into()));

rust-payjoin/payjoin-cli/src/app/v1.rs

Line 318 in 37d255e

.map_err(|e| Error::Server(e.into()))?

rust-payjoin/payjoin-cli/src/app/v2.rs

Line 345 in 37d255e

.map(|res| Psbt::from_str(&res.psbt).map_err(|e| Error::Server(e.into())))

It would be good to review usage of crate::receive::Error and improve error nesting generally.

[DanGould]

Rather than having v2 wrapping the v1 logic, might it make more sense to have some core PSBT mutation receive logic that gets wrapped by the network-specific details of either v2 and v1 protocols? The v1 module could handle encoding the errors for HTTP (as is done now, since Error::Server is ~a 5xx error and Error::BadRequest is 4xx), and the inner error could be made which don't make as much sense in the context of v2 where the client sender speak HTTP directly to the receiver, and inner typestate error could be made to be generic, but not over the irrelevant HTTP error abstraction.