Use precmd() method instead of adding our own layer of preprocessing

FIX: Unable to connect on Debian testing

Lasercutting: Relative moves with the jog dial should move with G0 instead of G1

This simple workaround should simply ignore initial layers with little
to no movements, and thus avoid M190/M109 to destroy ETA prediction.

basetrans was not updated before computing the real coordinates, thus
leading to invalid rectangles

This is needed to get sdlisting updated before checking it in the echo
condition.

get_eta could be called from statuschecker thread after self.p.printing
was set to True but before startcb was called in print thread, thus
leaving compute_eta uninitialized