add support for file cache from remote object store

[djouallah]

modern open query engines are getting so fast that now the bottleneck is basically remote storage bandwidth, you can see it from this chart, the biggest improvement to duckdb was adding remote cache

[shehabgamin]

@djouallah Sail helped with implementing a Parquet metadata cache in DataFusion, and this will be available starting with the DataFusion 50 release.

By file cache, do you mean something like that, or are you thinking of literally caching the Parquet data?

[djouallah]

@shehabgamin literally cache the columns and row groups used by the query either inside the vm ssd or in RAM, but parquet metadata and specially the list of files belonging to a particular delta version is more critical as a first step