Regressions in firmware retract

[richfelker]

Idempotency of the G10 and G11 commands is an important/key property of firmware-retract. It's what makes it so your start gcode can end in a retracted state (so it doesn't ooze and lose priming) even if your slicer (cough Cura) insists on performing its own retraction before the initial travel move, without getting a double-retract.

Maintaining statefulness of retraction after one print ends and before another begins is also important to keep filament retracted at the right point not to ooze or waste excess material priming the next print.

The changes in #83 make the former spam warnings, and completely break the latter by clearing retraction state after each print. They should be reverted and the system for introducing Z-hop re-thought not to break existing firmware-retraction semantics.

[rogerlz]

Thanks for the feedback. We are working on making the features from #83 optional and returning the default behaviour for users that prefer so until we can work on a better solution

#206