Verify F2 and F4 by using a decent block size

Fixes around F4

Use pkg-config to find libusb

CoreIF for F2/F4 is not different, use CPUID to distinguish. F4 errata seems to give wrong workaround

Allow to write to ram and start loaded programs. 

that case. Report advance in half-word writing. Setting stack and PC give
access error so automatic start fails too and hardware needs to be hard-reset
to start

by consecutive page erase
Allow to erase the device with the flash

SP value and the Program counter with the reset value from the Vector
table. Then run reset function

behaviour with L1 and F2

Conflicts:
	gdbserver/Makefile

As write_buffer_to_sram() can now handles arbitrary size, use flash_pgsz chunks for writing F1 and reduce output

moved the device parameter loading at the end of initialization for V1 devices- fixed the problem of non working V1 devices