Frame: decode payload length to return early

[678098]

A small optimization for partial Frames.
I have a feeling though that it makes things a bit more complicated, especially addressing buffer via exact offsets like buffer[3].
Might be good to add const size_t constants for these offsets.

Replacement for PR with Verified status

[willhoy]

looking at this again, I'm not 100% sure we don't rely on the type, channel being populated on a partially returned frame - @adamncasey do you have any thoughts here.

[adamncasey]

I also wasn't sure on this change but your question made me look in a few places which convinced me:

https://github.com/bloomberg/rmqcpp/blob/main/src/rmq/rmqio/rmqio_decoder.cpp#L55

If we return PARTIAL we never look at the (possibly) populated frame object.

https://github.com/bloomberg/rmqcpp/blob/main/src/tests/rmqio/rmqio_decoder.t.cpp#L71 this test covers this behaviour

[678098]

I believe I reviewed Frame usage when I was making this change, at least rmqcpp itself doesn't rely on fields in a partial Frame.

Another thing, we have a partial Frame early return on line 66 anyway:

rmqcpp/src/rmq/rmqamqpt/rmqamqpt_frame.cpp

Lines 59 to 67 in 98a7e38

	Frame::ReturnCode Frame::decode(Frame* frame,
	bsl::size_t* readBytes,
	bsl::size_t* missingBytes,
	const uint8_t* buffer,
	bsl::size_t bufferLen)
	{
	if (bufferLen < frameOverhead()) {
	return Frame::PARTIAL;
	}

when no fields are updated. This means, that the old version of the code has a strange situation: PARTIAL means that some fields might be or might be not initialized. And we still don't have a way to know which fields are initialized in this case.
That's the reason why not initializing fields for partial Frame in all situation makes it more clear.