Avoid nullptr dereference on bad rcmap value entry

[peterbarker]

Setting a bad rcmap entry doesn't just hard-fault the autopilot but ensures you have to reset all of your parameters before you have a working autopilot again.

This has been the case for a very long time. But perhaps we should fix it?

Checking the library return values for nullptr doesn't help if we have already persisted the parameters - next time you boot you'll not populate e.g. channel_roll with a valid pointer and will then segfault. So we use a "dummy channel" instead.

A new record for shortest autotest.

[peterbarker]

I've added an additional patch to MSP to avoid a similar problem.

There are almost certainly more in the codebase, but I want to know this is an acceptable solution before patching them all.

[peterbarker]

Problems around optimisation again on Pixhawk1-1M-bdshot; elf-diff shows many small increases in random methods

Board                    AP_Periph  blimp  bootloader  copter  heli  iofirmware  plane  rover  sub
CubeOrange-periph-heavy  *                                                                     
Durandal                            -16    *           8       -56               -256   64     72
Hitec-Airspeed           *                                                                     
KakuteH7-bdshot                     40     *           64      40                56     208    56
MatekF405                           16     *           8       32                176    0      80
Pixhawk1-1M-bdshot                  56                 344     384               280    48     32
f103-QiotekPeriph        *                                                                     
f303-Universal           *                                                                     
iomcu                                                                *                         
revo-mini                           32     *           16      136               112    48     56
skyviper-v2450                                         40                                      

[Ryanf55]

Suggested change

	// the library gaurantees that these are non-nullptr:
	// the library guarantees that these are non-nullptr: