This change only sets the "sensor" bits of VPChannelID when initialising with no value. The old version also set all of the bits and due to a mistake in !645 (merged) this means the pointToChannel function in VPSensor returned a VPChannelID where the orx & ory values were always set (fraction of channel in x and y). This was causing problems for the VP associators, and meant the linker tables were not correct. This fixes the issue raised in Lbcom!812 (merged)
Without this fix there will be issues with both linkers and VPClusters being offset by a fraction of a pixel using Retina Clusters.
