Blender Git Loki

Cleanup + Optimisations for GP Sculpt Multiframe

* Only initialise the falloff curve's internal data once per run,
instead of once per layer. We can probably even make this only happen
once for each run of the operator, as it shouldn't be possible to change
the curve in the meantime (requiring the data to be recalculated)

* Save off the multiframe falloff option into a variable instead of
doing the full check each time.

* Only calculate "diff_mat" once for each layer, instead of once for
each GP frame for multiedit. (Note: We weren't taking the frame numbers
into account anyway, so it doesn't matter to pull this out. However,
technically, if the layer was parented, we'd actually need to have been
recalculating the parents transforms for each of the multiedit frames
to get the real intended effects)