Removed inefficient LayerFilter check

[lossyrob]

The non-overlapping check took upwards of 2 minutes in a case where I was querying a zoom 12 512x512 GlobalLayout for the state of Nevada. This is a nice thing to ensure, but not at that price. Make a note in the scaladocs for the implementor's responsibility to ensure non-overlapping KeyBounds.

[lossyrob]

@echeipesh need advice here. That check really slows things down unacceptably in the workshop. But the failing test reminds me why it’s there (hint, it’s for the or that I doubt has been used in anger once). Is there some clever way we can get around having to do the costly check without allowing users to potentially do invalid queries (which I believe pulled back duplicate tiles)?

[echeipesh]

Actually this should be fine now. I think this check existed before we started using MergeQueue on the RDDReaders: https://github.com/locationtech/geotrellis/blob/master/s3/src/main/scala/geotrellis/spark/io/s3/S3RDDReader.scala#L58

Any intersecting ranges should be taken care of there.