DroneCAN: update libcanard

[tridge]

this reduces CPU usage in the hot frame accept path using hash on the message ID into a set of message lists

this is the key change: dronecan/libcanard#81

I profiled the change using SYS_STATUS.load while I had another flight controller on the bus sending a high rate of ESC RawCommand messages.
Here is the result on a CubeOrange with current master copter:

and here it is with this PR:

this shows roughly 10% reduction in overall CPU usage.
Note that this is a (deliberately) extreme case. We don't process RawCommand in ArduPilot main firmware, so we need to traverse all the way to the end of the linked list to determine that we shouldn't accept it, plus I setup the other flight controller with SCHED_LOOP_RATE=800 and as an OctoQuad, so over 2400 frames/sec of CAN frames we need to reject
The gain in CPU in more normal operation will be much smaller - maybe 2 to 3%? I setup the extreme case as it is quite a noisy thing to measure so much easier when you push the CAN bus hard

[tpwrules]

Curious if you've done any particular profiling?

[tridge]

Curious if you've done any particular profiling?

I've added some profiling data to the PR

[tpwrules]

This looks reasonable but I will try to test on hardware this weekend.

When I did testing I used a slightly tweaked --enable-stats and @SYS/threads.txt.

[tpwrules]

Looks good to me, I tested on CubeOrange it did not immediately break everything.

I want to get this in this PR too just to make bisecting easier and reduce the amount of submodule churn: dronecan/dronecan_dsdlc#35 (it's tested and ready to go in)

[tridge]

I want to get this in this PR

done