Tracking bug for metadata-only-import from S3 or other hosted storage

[olsen232]

A yet-to-be-fully-designed flow that creates a tile-based Kart dataset that is "backed" in some sense by the original data on S3 (or similar), rather than being necessarily stored in the local Kart repo or available-on-demand from the remote Kart repo. The storage format will be almost identical to a regular tile-based Kart dataset, except that extra information will be stored about the original location where the tile was found, so it can be fetched from there if necessary.

[olsen232]

Still TODO:

• Add a per-dataset "no checkout" config option
  • Expose which datasets are non-checkout in kart status
• Don't fetch entire PC tiles just to extract metadata
  • Update to PDAL 2.6 (see Upgrade PDAL to 2.6.0 - schema extraction changes #924)
• Fix kart lfs+ gc to clean files that were fetched from S3
• Make sure BYOD files are never pushed to LFS remote
• Rename byod-import commands
• add a remote-storage.json meta-item that allows for S3 fetches to be redirected
• add config options that allows for S3 fetches to be redirected