add support for a "debug" mode, or just make error handling more robust

[ralexstokes]

one thing I have noticed while using the Client type is that failures are annoying to diagnose

the API is strongly-typed and so if URL endpoints and message types are correct, things work well

if they aren't, it is usually hard to know what went wrong

options:

• add some type of "debug" flag inside the Client that exposes the information we need (e.g. which endpoint, what does the HTTP message look like as text, what does the response look like as text)
• more tolerant response parsing (e.g. try to parse within the realm we expect but if that fails, then make an Error variant that dumps the response status code and body (if any))

this last route actually sounds better (even if ultimately we do need the former)