[gnea/grbl-Mega PR#68] Fixes gremlin where changing step and direction inversion masks made a mess

This PR fixes a gremlin where changing the stepper step and direction inversion masks didn’t work correctly because stepportinvertmask[] and dirportinvertmask[] (new with the RAMPS modifications) weren’t reset to 0 when they should have been (when unset).

This made initial setup more difficult due to likely desynchronization between settings.stepinvertmask and stepportinvertmask[] and between settings.dirinvertmask and dirportinvertmask[] with a change to $2 or $3, where grbl would ultimately behave as though there was always an inversion.

(Also, while I might have used a ternary to fix this, the way I did it seemed most consistent with grbl style. Please advise if you’d rather a ternary – easy fix.)