I/O requests should have separate timeout

Like OpenDAL does, https://opendal.apache.org/docs/rust/opendal/layers/struct.TimeoutLayer.html

`object_store` should have in addition to https://docs.rs/object_store/latest/object_store/client/struct.ClientOptions.html#method.with_timeout

a `with_io_timeout` or similar setting that differentiates between "fast" requests (DELETE, LIST...) and requests that performs bandwidth intensive I/O operations.

In the current state, a user with a slow connection will always hit timeouts even though they are getting data through.

Start with the ClientOptions::with_timeout API linked in the issue and compare OpenDAL's TimeoutLayer behavior. Trace how object_store applies request timeouts, then define and verify separate handling for fast operations such as DELETE and LIST versus bandwidth-intensive I/O operations.

Written by the indexing model from the issue text.