DroneCAN: fix dangerous bug for twin-motor Planes

[robertlong13]

Without this, twin motor planes with DroneCAN ESCs need to set a
dummy throttle channel for the scale_esc_to_unity to work correctly.

Worse yet, if you don't do this, your ESCs will receive a max throttle command while disarmed as soon as the safety switch is disengaged.

The first commit prevents the max-throttle symptom when esc scaling is unconfigured or misconfigured (costs 16 bytes). The second commit allows the left/right throttle servos to be used to configure esc scaling (costs 24 bytes).

Params to reproduce this bug on CubeOrange, disengage the safety switch and look at RawCommand coming out of the autopilot in DroneCAN GUI

CAN_D1_UC_ESC_BM 3
CAN_P1_DRIVER 1
CAN_P2_DRIVER 1
SERIAL6_PROTOCOL 2
SERVO1_FUNCTION 74
SERVO1_MAX 2000
SERVO1_MIN 1000
SERVO2_FUNCTION 73
SERVO2_MAX 2000
SERVO2_MIN 1000
SERVO3_FUNCTION 0
SERVO4_FUNCTION 0

[IamPete1]

The whole scale to unity thing really should be removed at somepoint. It will cause issues if you were to set different ranges on your throttles, or PWM_MIN/PWM_MAX for quadplane.

We should calculate the scaled output per channel.

[tridge]

I'd like to AP_HAL/RCOutput.h change to init the min/max to 1000/2000

[robertlong13]

Finally got around to the testing to confirm that I haven't inadvertently messed with the math:

Params, all default except:

CAN_D1_UC_ESC_BM,4 # Servo 3 (throttle)
CAN_D1_UC_ESC_OF,2 # Offset by 2 (so only 3 comes in)
CAN_D1_UC_ESC_RV,4 # Servo 3 is reversable (turned this on and off to test both)
CAN_P1_DRIVER,1
CAN_P2_DRIVER,1
SERIAL6_PROTOCOL,2 # So I could have Planner and DroneCAN GUI connected simultaneously

Armed in manual and throttled up using vjoy through Mission Planner and looked at the RawCommand messages using DroneCAN GUI

Tested in/outs:
(edited: had the headers reversed)

RC3 %	CMD (non-reversable)	CMD (reversable)
0	0	-8191
25	2047	-4095
50	4095	0
75	6143	4095
100	8191	8191

Also, I've made the change the Tridge requested.

And of course I also tested that this PR does still in fact fix the bug.

[tridge]

@robertlong13 I do hope that table has swapped column headers for reversible/non-reversible?

[robertlong13]

@robertlong13 I do hope that table has swapped column headers for reversible/non-reversible?

Yup, swapped headers. Sorry. Editing the comment now