MAVLink: Generate correct RC channel count

[MaEtUgR]

Introduction
According to the message definition https://mavlink.io/en/messages/common.html#RC_CHANNELS_OVERRIDE
channels 1-8 should be ignored when they contain the value UINT16_MAX and the MAVLink 2 extension channels 9-18 should be ignored if they are either UINT16_MAX or zero.

Problem
Currently, the channel count is always 18 which doesn't reflect this definition. Hence I'm implementing the definition.

Solution
We loop from the last to the first channel and if one of them doesn't satisfy the conditions to be ignored we set the channel count to that number. Hence if the message is full of UINT16_MAX's the channel count stays 0, if the message is full of zeroes the count will be 8 because of the different definition in the extension and everything in between gets checked if it's ignored.

Test data
We use this implementation for half a year now on the H520.

Additional Context
mavlink/mavlink#855
#10535

[potaito]

Makes sense 👍

[LorenzMeier]

CI fail:

+ px4-px4_sitl_default-v1.9.0-alpha-226-gcb1014e/px4/Tools/ecl_ekf/process_logdata_ekf.py .ros/log/2019-01-15/08_42_44.ulg
<type 'exceptions.IndexError'> (estimator_status): list index out of range
<type 'exceptions.IndexError'> (ekf2_innovations): list index out of range
Using test criteria loaded from /tmp/jenkins/workspace/L_tests_mavlink-rc-channel-count/px4-px4_sitl_default-v1.9.0-alpha-226-gcb1014e/px4/Tools/ecl_ekf/check_level_dict.csv
Traceback (most recent call last):
  File "px4-px4_sitl_default-v1.9.0-alpha-226-gcb1014e/px4/Tools/ecl_ekf/process_logdata_ekf.py", line 83, in <module>
    late_start_early_ending=not args.no_sensor_safety_margin)
  File "/tmp/jenkins/workspace/L_tests_mavlink-rc-channel-count/px4-px4_sitl_default-v1.9.0-alpha-226-gcb1014e/px4/Tools/ecl_ekf/analyse_logdata_ekf.py", line 35, in analyse_ekf
    innov_time = 1e-6 * ekf2_innovations['timestamp']
KeyError: 'timestamp'

@priseborough Do we need to extend the EKF check tooling?

[LorenzMeier]

@MaEtUgR Could you please rebase? Thanks!

[MaEtUgR]

I rebased, addressed review comments and improved readability.

[dagar]

I still don't care for the magic number, but I don't see anything from mavlink we can use directly.