This PR demonstrates the known regressions of the recently merged material branch:
* Changing the transparency after setting color per face will reset them
* The result of boolean operations or compound doesn't inherit the colour of its input objects
* If colour is set per face to a boolean operaton object then saving and restoring the file causes weird rendering behaviour
because material binding is set to PER_PART but only a single colour is defined
* If a shape inside a part container has set colour per face then saving and restoring as STEP file causes weird rendering
behaviour for the same reason
* Shape binder or datum objects don't show the correct default shape colour