Changes to RRF for RTOS

[done] Include FreeRTOS library in the link library list and run RRF as a single FreeRTOS task
[done] Run Heat as a separate task
[done] Replace delay() by a version that calls FreeRTOS
[done] Have hsmciIdleFunc suspend the calling task pending an interrupt
[done] Use a mutex to protect fatfs from being reentered, preferably a separate mutex for each volume
[done] Use a mutex to protect the shared SPI subsystem from reentrant calls, and use it the filesystem and the temperature sensor reading system
[done] Use a mutex when allocating file handles
[done] Use a mutex to protect the find first/next buffer
[done] make getting file info thread safe
[done] buffer allocation needs to be thread safe
[done] tool list traversal to be thread safe
[done] make USB and serial out channels thread safe
[done] make message box output thread safe
[done except for mutexes on network and telnet] Make Platform::Message thread safe
[done] Get rid of shared scratchString
[done] Make interface between Network received data and GCodeBuffer thread safe, ready for running as separate tasks
[done] FileGCodeInput: do away with unnecessary local buffer
[done] Run Network as a separate task
Run Move as a separate task. Probably needs a mutex to protect calls into it, e.g. babystepping and pause.
Use DMA for SharedSPI transfers, suspend the calling task while waiting on a transfer
Do a SAM4S version of FreeRTOS, then do a SAM4S_RTOS build of RRF
Stop allocating file buffers on the stack?
Find maximum stack sizes and optimise stacks
Run all network responders as separate tasks? Would need mutex on network interface access

Bugs

[done] IAP fails to update firmware
[done I think] M122 output is too big for the available buffers, unless the Network task reads the output fast enough
[probably related to above] Duet WiFi gets into a state in which it returns empty JSON messages after sending M122 from DWC. Output buffer starvation?
[done] Disconnects when getting file info forCandyBowlx8comments.gcode when running RTOS build
[seems better now] File upload speed is a little lower when using RTOS - may be fixed when network is a separate task
[seems ok now, was probably main stack overslow] Report that Pause works, but then hangs
[done] Heater tuning crashes the firmware
[believed fixed] DWC disconnects randomly with JSON parse errors
 

Deferred work

Use our own vsnprintf, and stop using sscanf. Should save a significant amount of stack space.
Check whether we use any time conversion functoins that use _reent
SAME70 uses a different memory model i.e. not strongly-ordered. Does this have any consequences for the code so far?
Are there any aspects of Platform that are not thread-safe? e.g. GCodes setting multiple interdependent variables while network reports on them?
Use a mutex to protect HSMCI from reentrant calls (only needed if/when we support more than one HSMCI volume, which is unlikely but possoible on the SAME70)