AsyncPayloadProcessor owned payloads are released too early

[RobGeada]

Issue:

AsyncPayloadProcessor sometimes sends null response payloads, despite the original infer caller getting a valid output back to the calling process.

Triggers

If the payloads are processed instantly by the AsyncProcessor, the majority (but not all) responses arrive intact. If the AsyncProcessor is slow, the majority of responses arrive with 'null' data.

Investigation

It looks like the response bytebufs are read too early somwehere, causing their readerIndex to equal their capacity at processing time. This read occurs somewhere between their addition to the queue in the AsyncPayloadProcessor but before payloads.take() is called. A hacky patch is to call byteBuf = byteBuf.readerIndex(0); as the first line in RemotePayloadProcessor.encodeBinaryToString(), to reset their reader index, and indeed this prevents the issue from arising.

[RobGeada]

This seems to the underlying cause of trustyai-explainability/trustyai-explainability#368