Improve RTCM3 parsing performance

[alexcekay]

Solved Problem

This is the first follow up to: PX4/PX4-Autopilot#24491. The NAV + RTCM3 parser can block each other as both parsers are integrated into the same state machine. This can cause significant delay until the parser gets in sync again.

As can be seen in the attached screenshots, the remaining timeout is in general higher now and the deepest drop is not as low as before. This was the worst case scenario of many tests.

Worst without change

Worst with change

Solution

• Forward input data to both NAV + RTCM3 parser and keep their states separate, so they can not block each other
• Use CRC24 to verify if RTCM3 data is valid before forwarding it to the callback
• When either NAV or RTCM3 parsing is successful, reset both parsers to start sync again

I would like to keep the increased timeout until all remaining drops are removed, which will be done in follow-up PRs.

Testing

• Tested with and without RTK on the bench
• Checked gps_dump data to ensure PPK still works

[alexcekay]

Hi @niklaut, @bkueng,
could you guys have a look at this PR? Thanks.

[bkueng]

I would reset here in any case
if (_rtcm_parsing) {
to be independent from whether a message was handled or not.

And then move the block from below to above:

	if (_rtcm_parsing) {
		if (_rtcm_parsing->addByte(b)) {
			gotRTCMMessage(_rtcm_parsing->message(), _rtcm_parsing->messageLength());
			decodeInit();
			_rtcm_parsing->reset();
			return 0;
		}
	}

[alexcekay]

Thanks for the input. I updated it.

[bkueng]

Looks good, nice work