use error value for bad URI so custom error handler could treat it special

[jhump]

When a custom ProtoErrorHandlerFunc is used, it is given an error with Unimplemented gRPC code for cases where the server gets a request for an unrecognized URI path.

When no such handler is in place, it defaults to a "404 Not Found" error, which is what basically any REST client actually expect for such an error.

Unfortunately, there is not currently any way to distinguish between "the server received an unrecognized URI path" and an actual "Unimplemented" error returned by a gRPC server handler. So it is not possible for a custom handler to turn those into "404 NotFound" errors, as a REST client would actually expect.

This change hoists the error out into an exported error value: ErrUnknownURI. That way, a custom error handler can check if err == runtime.ErrUnknownURI and choose to return a 404 for those errors (instead of "501 Not Implemented", which can be used for other errors with an "Unimplemented" gRPC code).

[codecov-io]

Codecov Report

Merging #932 into master will increase coverage by 0.28%.
The diff coverage is 33.33%.

@@            Coverage Diff             @@
##           master     #932      +/-   ##
==========================================
+ Coverage   53.53%   53.81%   +0.28%     
==========================================
  Files          40       40              
  Lines        3964     3954      -10     
==========================================
+ Hits         2122     2128       +6     
+ Misses       1645     1630      -15     
+ Partials      197      196       -1

Impacted Files	Coverage Δ
runtime/mux.go	54.16% <33.33%> (+9.28%)	⬆️
runtime/proto_errors.go	0% <0%> (-34.89%)	⬇️
runtime/errors.go	47.36% <0%> (+2.63%)	⬆️
runtime/handler.go	40.62% <0%> (+5.28%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f005bec...fcb8708. Read the comment docs.

[achew22]

This seems eminently reasonable. Would you be willing to add a test or two to confirm the differentiation between a 404 not found and a 404 unimplemented?

[jhump]

@achew22, I added tests. PTAL.

[johanbrandhorst]

Thanks!