Make coalesce_ranges and collect_bytes available for crate users

[sumerman]

Which issue does this PR close?

I didn't create an issue

Rationale for this change

Both coalesce_ranges and collect_bytes are useful for AsyncFileReader implementations that don't quite fit object_store interface, but address the same fundamental problem — fetching data from a remote location.

What changes are included in this PR?

coalesce_ranges, collect_bytes and OBJECT_STORE_COALESCE_DEFAULT are made pub use

Are there any user-facing changes?

[tustvold]

#4786 should resolve the clippy failures

[sumerman]

#4786 should resolve the clippy failures

Thank you! I've rebased my PR

[sumerman]

Another change, that I believe would be useful for non-object_store uses, is switching to std::result::Result type for coalesce_ranges and allowing errors that aren't object_store::Error. WDYT?

[tustvold]

Adding the error as a type parameter doesn't seem like it would be particularly controversial / complex. I am somewhat curious what you are intending to use this for though?

[sumerman]

Adding the error as a type parameter doesn't seem like it would be particularly controversial / complex.

👍🏼 I'll send it as another PR

I am somewhat curious what you are intending to use this for though?

We have an AsyncFileReader implementation that, firstly relies on in-process metadata caching, secondly manages its own tokio runtime (we have to be careful with the runtime as we run inside PostgreSQL). So, we can either sidestep object_store or we would have to build both an object_store impl that is compliant with our runtime needs and AsyncFileReader that can use the metadata cache, but delegates the rest to the object_store.

[tustvold]

Makes sense, I do hope to eventually flesh out a first-party caching story, but I don't have anything to show for it as this time 😅