python: Expose laz reader in Python bindings

After https://github.com/apache/sedona-db/pull/471, we have an laz reader thanks to @b4l!

I think packaging pointcloud readers and future possible writers should be done in a separate Python package...even though it could be done by building everything into the main Python package, at some point we have to start separating things and DataFusion provides some great utilities to do this.

I think the initial version of this could expose a listing table (table provider) or user-defined table function and leverage the existing DataFusion FFI utilities. I think these went through some significant changes in DataFusion 52, so it may be best to wait until then to ensure spatial predicates can make their way through for pruning.

Start by reviewing the existing laz reader from PR #471 and DataFusion’s FFI utilities, focusing on the proposed separate Python package and table-provider or user-defined table function. Done means agreeing on and implementing a Python-facing listing table or UDTF that exposes the laz reader and supports spatial-predicate pruning; no source files or tests are named.

Written by the indexing model from the issue text.