Standardize assumptions about sensor pre-processing

[hidmic]

While working on #491, we decided that it wasn't Beluga's responsibility to pre-process point clouds in application specific ways, potentially taking a performance hit (point clouds can grow large). Alas, we do filter laser scans as a result of our good ol' AMCL heritage.

We need to reconcile and standardize assumptions so that it isn't confusing on users or developers.