don't escape paths twice when using a schema registry

[huttotw]

In the case where we had a subject like my/deeply/nested/message.proto it was being doubly escaped like:

/subjects/my%252Fdeeply%252Fnested%252Fmessage.proto/versions/1

when it should be:

/subjects/my%2Fdeeply%2Fnested%2Fmessage.proto/versions/1

Note that by default Go will unescape in the Path field of url.URL, to access the raw path you can use RawPath. Not sure if we should change this test func to use the RawPath or not?

[Jeffail]

Hey @huttotw, is there a formal spec for how a schema registry server expects these paths to be formed? We've had issues with this before: #2190 but I find it very difficult to get this "right" when we're having to guess based on different implementations of the server.

[huttotw]

@Jeffail thanks for the quick response!

I'm not certain that the bug was fixed in #2190, I tried to reproduce the scenario here. Let me know what you think!

I'm referencing Confluent's docs here, but the only conclusion I can draw is that subject should be treated as an opaque identifier:

https://docs.confluent.io/platform/current/schema-registry/develop/api.html#get--subjects-(string-%20subject)-versions-(versionId-%20version)

Nevertheless, I don't think we want to escape twice; that seems a bit unexpected from all sides. It looks like we should be using the JoinPath method instead of manually setting the Path based on a JoinPath result, there are some other things happening under the hood it looks like when Path gets set via JoinPath.