ADR: Why Polars?

Context: We need a library to perform data transformations (ETL).

Options Considered: Pandas, PySpark, Polars.

Decision: Polars.

Rationale:

PySpark is overkill (too much overhead) for single-node processing.
Pandas is slow and memory-inefficient for large datasets.
Polars offers blazing fast multi-threaded execution and a clean, expressive API.