Representing caching / multi-cloud data

[cholmes]

Imagery may often live in multiple places. Landsat data lives on USGS servers, AWS and Google Cloud (on Earth Engine). It is widely used and useful to be close to algorithms that can process it. It may often make sense to copy data from one bucket to another, to be closer to other, larger data.

We should have a way to represent that an image is a direct copy of another one. This could allow clients to realize that there may be a better location for it. And indeed catalogs could link to the various locations a data set is stored. #13 discusses some options for derived data. Perhaps a cached copy is the simplest derived data case - it's an identity operation with a different location online.