Add flag for smoothness prior on intrinsics

[johnwlambert]

We have one method for shared intrinsics currently (via separated intrinsics and poses): #65
This method slows down BA, presumably due to increased fill-in.

We propose to have a second option as a flag, which allows shared intrinsics, via a smoothness prior between the calibrations. In this way, the graph can stay sparser via the Markov chain of calibrations.

This may be preferable when the camera heats up on a drone, for example.