Try to have sensible setup/hold times by evenntually duplicated or
logically useless port commands.
Loading code to RAM on a STM32L476 got up from 46 to 83 kB/sec.
This adds a new function to the internal target interface
to allow the target to get control before reset is complete
so that it can do any additional work. On this target there
is a proprietary internal bit that has to be reset in some
cases to allow the core to continue operating.
