feat: integrate object_store for read/write with pyarrow

[roeap]

Description

This PR embraces the object store crate also on the python side and at least for the current test base supports reading and writing using the dataset and other pyarrow APIs. Thanks to @wjones127 's mutipart upload in object_store, implementing the write functionality was actually quite straigt forward. We now implement ObjectInputFile and ObjectOutputStream, which - if wrapped in pyarrow.PythonFile will work with the arrow ecosystem (so far ;))

There was on bigger and breaking deign decision, but I hope people agree :).

Essatially I just accepted, that working exclusively with the relative delta paths makes life much more convenient.. As such rather then adding and removing paths prefixes all the time, I thought it would be reasonable to as users to wrap their own filesystems in a pyarrow.fs.SubTreeFileSystems, which points at the table root..

import pyarrow.fs as fs
from deltalake import DeltaTable

path = "<path/to/table>"
filesystem = fs.SubTreeFileSystem(path, fs.LocalFileSystem())

dt = DeltaTable(path)
ds = dt.to_pyarrow_dataset(filesystem=filesystem)

Of course the hope is to eventually be at comparable (or higher :)) performance then the c++ file systems. Then there would be little reason (I guess) to still provide a "custom" file system at all.

Related Issue(s)

closes (at least to some degree) #570
closes #574
closes #696
closes #689
towards #542

Documentation

[houqp]

LGTM, great work!