Do DataFrame operations under the hood using Polars for q.cells()

[luke-quadratic]

Pyodide added Polars support recently, should enable a huge performance uptick for Python if we decide to use it: Pyodide PR for Polars

A few possibilities:

• q.cells() - use Polars under the hood for operations and convert to DataFrame in last step by default
• have an option on q.cells() to allow Polars instead of Pandas to give users a huge performance uptick (enabling the option would enforce Polars DataFrames to that user's initial return type)

Regardless of whether we add the Polars instead of Pandas option as a parameter in q.cells() we can use Polars in the intermediate steps.

To be explored during upcoming perf work.