`PunctureTracker` interpolation is inefficient

[mirenradia]

Currently, all MPI ranks request interpolation of the shift in the puncture tracker but this is unnecessary as only rank 0 writes the output. Note that other ranks may need information on the location of the punctures (e.g. for tagging) so the best approach would be for only rank 0 to request the interpolation of the shift and then broadcast the calculated position to all other ranks.

Note that is similar to what was done in #147 for SurfaceExtraction so an alternative approach would be to create an interface to the AMRInterpolator that only interpolates onto rank 0?