stmhal: Improve efficiency of SysTick IRQ and HAL_Delay.

SysTick IRQ now increases millisecond counter directly (ie without
calling HAL_IncTick).  Provide our own version of HAL_Delay that does a
wfi while waiting.  This more than halves power consumption when running
a loop containing a pyb.delay call.  It used to be like this, but new
version of HAL library regressed this feature.