This provides a clean ROS interface for external projects which is
independent from the px4 CLI interface.

It can be used with:
```
	<include file="$(find px4)/launch/px4.launch">
	</include>
```

without this tasks will be switched all the time and the drone starts driftig

 - a unity build is a single compilation unit per module

   hardware detection was overwriting the CAN2 GPIO setting

The -P argument seems not to be supported on OSX

The Simple HW detection was detecting the
  Pixhack V3 as a mini.

  This commit adds further discrimination, based on
  the fact the Pixhack V3 left VDD_5V_PERIPH_EN
  floating. Where a mini ,V2, V3 HW has it connected
  to the active low enable of the VDD_5V_PERIPH and
  VDD_5V_HIGHPOWER via a 10K pull down.

  The detection enables the 40K pull up and samples
  the pin. This reads back as a logical one on the
  Pixhack V3, and a logical zero on the Mini.

  Since the float is applied some 3.3 Ms post pin setting
  at reset the 10 us should leave the 5 Volt supplies in
  the on state because of the 10K pull down.
  This is assuming typical 40K pullup 10K pull down
  This results in typical 0.66V (.875V max) on the /EN pin
  which should not be detected as a high on the TPS2041.

  On a FMU without a px4io (FMUv4, NXPhlite etc), this code should
  not be run.

   One a 32.768 is added (RC16) to the board this will keep time.

   The fix ix in nuttx. This is just the defconfig changes to use
   it.

   There were 2 problems. The first was that the interrupt did
   test DINT and raise a completion events. But since DINT is
   just an indication of DMA completion, TC is a valid way to
   determine that the transfer is complete.

   The second problem is that Software Reset For DAT Line
   SDHC_SYSCTL[RSTD] clears the bits 24-0 in SDHC_PROTO
   this looses the wide bus setting DTW