Consider making pandas an optional dependency

[FBruzzesi]

Description

Currently pandas is a hard requirement for the library, however not necessarily used.

By delaying pandas import in src/databricks/sql/client.py from the top of the file to _convert_arrow_table method after the if self.connection.disable_pandas is True: condition it would be possible to make pandas an optional dependency.

I was able to run all the tests with this change flawlessly, as well as uninstalling pandas and then run the following:

import polars as pl
from databricks.sql.client import Connection

credentials = ...
query = ...

conn = Connection(
    server_hostname=credentials["server_hostname"],
    http_path=credentials["http_path"],
    access_token=credentials["access_token"],
)

dframe = pl.read_database(query, conn)

without any issue