No error handling scheme

[mikemccracken]

Some functions such as tag_new, tag_delete and tag_machine return a bool success value for some cases (but will let any exceptions from the REST layer through), but higher-level functions that use them ignore the return values - e.g. tag_name and tag_fpi. Those will fail silently in event of an error, or just pass exceptions up.

We should handle errors the same way across all functions and surface them to the client somehow - return values or exceptions, either way.