STM32 F401:  Only 3 USARTS, but need to set STM32_NUSARTS to six because they are not numbered sequentially

wdog.h does not contain any application interface, only internal OS interface.  Further, it is non-standard.  Move wdog.h from include/ to include/nuttx. For the same reason, move the description of the watchdog timer interfaces from the Users Guide to the Porting Guide.

NSH link management now works!  The last fix was to the Ethernet drivers:  They cannot disable clocking to the Ethernet blok on ifdown.  Otherwise, we cannot communicate with the PHY

Change the way PHY interrupts work: disable automatically.  Then we have to re-subscribe each time after the interrupt fires

More of the PHY event notification logic change:  Fix some compile errors when full feature is enabled; Add some missing ioctol logic

Modified to support the change to the network ioctl signature changes.  Also add support for new ioctl to setup PHY event notifications.

SAMA5 free-running timer: Add support for a free-running timer wrapper around the low-level timer/counter logic.

SAMA5 timer/counter: Add support for a one-shot timer wrapper around the low-level timer/counter logic.  This also involved several changes that rippled into the ADC driver (untested).

SAMA5 T/C: Can now handle non-constant BOARD_MCK_FREQUENCY. Also now supports methods to attach user interrupt handlers

Remove os_internal.h it has been replace by several new header files under sched/.  There have been some sneak inclusion paths via os_internal.h, so expect a few compilation errors for some architectures

Change all time conversions.  Yech.  New timer units in microseconds breaks all existing logic that used milliseconds in the conversions.  Something likely got broken doing this, probably because I confused a MSEC2TICK conversion with a TICK2MSEC conversion.  Also, the tickless OS no appears fully functional and passes the OS test on the simulator with no errors

Change CONFIG_MSEC_PER_TICK to CONFIG_USEC_PER_TICK.  This gives more options for system timers in general, but more importantly, let's us realize higher resolution for the case of CONFIG_SCHED_TICKLESS=y -- of course, at the risk of some new interger overvflow problems