Tests for inflator failure.

[Fischiii]

#8551

Added tests to reproduce the issue.

Added one test with the broken deflated message:
"inflater left with deflated bytes"

Added another test with a similar, but working message:
"inflater left with deflated bytes working message"

The first test results in the exact same behaviour as we see within okhttp when we discovered the issue: #8551

[yschimke]

@Fischiii a big thanks for doing this.

[yschimke]

OK, thanks for the repro. See Fischiii#1 that you can merge into your PR.

But

• I'm not clear if the server is sending additional bytes with per message deflate enabled.
  or
• OkHttp isn't correctly accounting for the additional bytes at the end of a message

Context takeover discussed here https://websockets.readthedocs.io/en/stable/topics/compression.html#compression-parameters

Final block discussed here https://datatracker.ietf.org/doc/html/rfc7692#section-7.2.1

@swankjesse do you remember the algorithm here?

[Fischiii]

@yschimke merged