Mavlink fixes & perf improvement

[bkueng]

• add optional disable_sharing flag to add_orb_subscription.
  It is a more generic solution for 532a970, and also enables it for vehicle_command_ack's.
• MavlinkOrbSubscription::update:
  • reorders operations, such that the most expensive one (orb_copy) is done only when really needed.
  • fix corner case: when the topic was not advertised yet, orb_stat() would fail and then update() was called, which succeeds for the first advertisement.
    In that case the timestamp was incorrectly set to 0 and true was returned. The next call would again return true, because the timestamp was updated, but the topic data was still the same.

Reduces CPU load by ~2% on a Pixracer.

[jdzerve]

This should be revisited.
The behavior has changed but description has not. Strictly speaking description was incorrect before this commit, too.
We are talking about MavlinkOrbSubscription::update(uint64_t *time, void *data)
In particular:

1. Description "If no data available data buffer will be filled with zeros." is/was not true.
   https://github.com/PX4/Firmware/blob/75f1ad36d1210560f00da8f4fa3bddbb3da9da16/src/modules/mavlink/mavlink_orb_subscription.h#L62

2. Description part "will still copy the data" is not true any more.
   The problem is that there is a code in MavlinkStreamSysStatus::send which actually relies on the documented behavior (copy even if not updated):
   https://github.com/PX4/Firmware/blob/75f1ad36d1210560f00da8f4fa3bddbb3da9da16/src/modules/mavlink/mavlink_messages.cpp#L571

The result is that sometimes SYS_STATUS message goes out with zeroes in onboard_control_sensors_* fields.

I could provide PR for this particular case but cannot do 100% fix as I'm not that familiar with code to know where else this behavior is relied upon.

[LorenzMeier]

Thanks for raising this!

[bkueng]

@jdzerve yes, please send a PR to update the comments and fix SYS_STATUS, thanks!
@eyeam3 are you aware of more places?

[dagar]

I also stumbled across this when updating the uORB subscription wrappers in mavlink for #11176. In that PR I restored the behavior as described in the comments (zeroing if no data, copy even if not updated) as a precaution, and because it was cheaper with the new interface.

That PR is going to take time to settle after v1.9.0 is released. I would suggest a quick fix for SYS_STATUS right now.

[jdzerve]

This is my fix:
jdzerve@4f59cf2
@dagar: Yours is little more involved, so I guess it's more appropriate.