Follow-ups of the SDK wrapper change

[nicholasjng]

A few things were left over from the SDK wrapper migration in #231.

• [x] Implementing file reads and writes with lakefs objects, see https://github.com/treeverse/lakeFS/blob/master/clients/python-wrapper/lakefs/object.py for details. (#240)
• [x] Implementing deletes via lakefs.Branch(...).delete_objects(*rpaths), both rm_file and rm. See https://github.com/treeverse/lakeFS/blob/af2394abd136a1c33b7dd2d3cff29d0fde40c822/clients/python-wrapper/lakefs/branch.py#L174 and the usage example in the docstring.
• [x] Implementing object copies as writes between branches if there's a nice API on the Branch class, otherwise leave as is. (#240)
• [ ] Find a nicer implementation of rev_parse for the demo than the one I made.
• [x] Implement a generic type ConcreteOrAbstract = T | Placeholder[T] (name TBD) and use it to improve ref typing on some of the transaction APIs. (#245)
• [x] Wrap LakeFSFile.details in the wrapped_api_call context instead of the super call in (get|put)_file. (#240)

We should also control all leftover lakefs_sdk imports to check if an equivalent name or better solution exists in the lakefs package, and if so, use that instead.

[nicholasjng]

That being said, do we even need a generic abstract value with wrapt if ininstance checks hold anyway for placeholders? That should relax the interface typing, so point 5) might not be necessary.

EDIT: If we go that route, we should implement a value getter that raises the runtime error on an unfilled placeholder that we had previously.

[nicholasjng]

I raised the remaining point 4 (better revision parsing) with treeverse in https://github.com/treeverse/lakeFS/issues/7232 - I consider this the appropriate place for the implementation (if it is happening), but we can roll our own anytime if they decide against it.