FR: handle and support a provided object version identifiers in data imports

[achtsnits]

background
LakeFS supports referencing external data through imports without copying it, but currently only references objects by path, without considering additional parameters like a versionId

note: while object locking is one way to enforce WORM principles on the external storage system, enabling bucket versioning provides a less intrusive mechanism to achieve immutability

proposal
if external data will be "just" referenced in LakeFS (e.g., for lineage, publishing, or sharing), it is critical to ensure that objects remain unchanged after import - versioned buckets provide a way to enforce this immutability and if LakeFS considers the versionId it guarantees this behavior.

handle

• during import
  consider a provided versionId, e.g., lakectl import --from 's3://<bucket>/test.txt?versionId=<version-id>' --to lakefs://<repo>/<branch>/test.txt

• retrieval
  API (direct & pre-signed URL generation) and S3-gateway should return the specified (and not latest) object version

side note: not letting LakeFS manage the object lifecycle will miss out on many of the valuable features LakeFS offers right out of the box (gc,...)...but that is another story