[CONTRIBUTION] Feature Parity: Implement missing `download_csv` and `get_latest_result_dataframe` in `AsyncDuneClient`

[nice-bills]

Currently, there is a functional disparity between DuneClient (sync) and AsyncDuneClient. Several high-utility methods available in the synchronous client are missing from the asynchronous implementation, specifically:

download_csv(): Essential for fetching large datasets efficiently as streams.
get_latest_result_dataframe(): A convenience method widely used by data scientists to fetch results directly into Pandas.
run_sql() (Plus feature): For executing raw SQL.

Users migrating to async workflows (e.g., using asyncio to run concurrent queries) currently lose access to these convenient methods. They are forced to either:

1. Manually implement the HTTP fetching/parsing logic.
2. Fall back to synchronous calls, blocking their event loop.

I would like to contribute a PR to implement these missing methods in AsyncDuneClient, ensuring they match the behavior of their synchronous counterparts in client.py but using aiohttp / async patterns.

Proposed Changes:

1. Implement download_csv in client_async.py using aiohttp streams.
2. Implement get_latest_result_dataframe (wrapping the CSV download).
3. Add relevant unit tests.

If this aligns with the project's roadmap, I am happy to submit a PR!