Back compatibility for vt.parse

[springmeyer]

After #419 is done the need for vt.parse() will be gone.

This is awesome, but we'll need to think through how to lessen the damage downstream.

My thinking is that:

• We should remove all usage of vt.parse from the node-mapnik tests
• We keep vt.parse() as a function but either:
  • make it do nothing except print a deprecation warning, or
  • change its behavior to be a validation function.

By validation I mean it could create a tile_datasource on the stack and make sure that works without throwing. That would allow usage like this to keep working the same: https://github.com/mapbox/mapbox-upload-validate/blob/7666dca170e0d35ad81280c2b2fb997c8852f72a/lib/validators/serialtiles.js#L48-L73. However I'm not yet convinced this is a good idea. There are a lot of forms of validation users might want for vector tiles and it likely makes more sense to think of this as a totally separate task. Users might want to know 1) is it a protobuf, 2) is it a valid vector tile at the byte/schema level ("parseable"), 3) is that vector tile made up of geometries that make sense (clipped and within the bounds of the tile) or was it improperly encoded if various odd ways...

[springmeyer]

change its behavior to be a validation function

For now I've kept it as a way to validate that the underlying buffer is parseable. Will revisit as I continue to work on #422