Discovery: do we want to publish the assembled Spark metadata client to Maven central

[talSofer]

Currently, we only publish the non-assembled Spark metadata client to Maven central. This task is to do some discovery and decide whether we want to publish the assembled jar too.

[arielshaqed]

We also do the exact opposite on lakeFSFS: there we publish assembled and plain JARs. At the very least we need to be consistent 😮

[arielshaqed]

In fact we publish (only) assembled metadata clients for GC (users need only add S3 / Azure hadoop plugins, depending on what they intend to use). Unassembled versions will be too hard for most users to get working.

Spark client is used almost exclusively for GC, which is generally performed by admins (or lakeFS.Cloud, of course).