Improvements, selective replacement for the spark and lakehouseing

[ZergRocks]

Integration between spark and dbt should be configured as a thirft server or session when running locally, and I think this can be to allocate more resources than necessary unless it requires multi-node processing.

I think with the dbt-duckdb and the built-in delta-rs or pyiceberg plug-in, we can dramatically reduce this weight in the composition of the aio directory docker files that you've constructed in ngods stocks

In addition, dbt-duckdb currently experimentally supports the iceberg external table option, and delta external table is also being developed by recent PR.

I'd like to ask for your opinion