Revert "drm_hwcomposer: Workaround for screen de-activating causing db845c regression"

This workaround actually leaves pipes in an unclean state on external displays which leads to planes bleeding and incorrect renderings in the case of overlays being assigned to external displays.

There is also a case to be made to not have board specific hacks in drm_hwc and we should fine a better solution if this is still a problem. We could consider only deactivating non-primary display if the new pipeline is set to null for instance.

This reverts commit 799e8c74.