Add AP_Filesystem API

[tridge]

This adds an AP_FIlesystem API for all filesystem operations. It cleans up the existing posix abstraction in ChibiOS, and will enable us to add automatic remount on operation failures
A new parameter LOG_FILE_TIMEOUT is added which gives the timeout in seconds for log writes. Setting this to 30 allows you to remove the card for 30s and when re-inserted logging will continue.
This is an alternative to #8639 from @peterbarker

Issues:

• adds about 1.7k to flash for unknown reason (fixed, is now smaller than master)
• add automatic remount on IO error

[tridge]

@WickedShell please check if the AP_Scripting changes are OK (also need to check it all works).
Note that this fixes the recursive opendir issue

[tridge]

this was flight tested today on a bixler

[WickedShell]

This breaks scripting on real boards. Works fine in SITL, but on a CubeBlack build the same script which previously worked now dies with unexpected symbol on the first line when parsing the script. I haven't gotten to investigate it yet, but it looks like it's probably reading corrupt data off the SD card. It's also notable that scripting is still implicitly using remove, and getting away with it, which is probably not correct with the AP_FS abstraction. The sample script I used was this:

function loop ()
  gcs:send_text(0, "LOOPING")
  file = io.open("test.lua", "r")
  gcs:send_text(0, file:read("*all"))
  file:close()
  return loop, 1000
end

file = io.open("test.lua", "a")
-- sets the default output file as test.lua
io.output(file)

-- appends a word test to the last line of the file
io.write("-- End of the test.lua file")

io.close(file)

return loop, 1000

[tridge]

@WickedShell ok, thanks. I'll fix this soon

[tridge]

@WickedShell I've now got it working with Lua scripts, and also now have a common FILE implementation between SITL and ChibiOS, so valgrind does useful checking of Lua IO calls

fixed

[davidbuzz]

I'm not sure why this change to the proximity sensor is included in the AP_Filesystem PR.? I haven't looked into the implications of the change from sscanf to strtol, and it's probably a harmless change, but thought I'd note it here anyway.

[davidbuzz]

I've read through all the code changes in the AP_Filesystem PR, and had just one comment that I added above. everything else looks like it's a pretty straight-forward relocation of the existing filesystem functions into the AP_FS:: namespace without too much to worry about.
... although it does touch a lot of files.

[tridge]

merged after discussion with mdb