add support for hf://buckets in httpfs

[cfahlgren1]

Hugging Face recently introduced Storage Buckets as mutable, non-versioned object storage on the Hub, addressable as hf://buckets/<namespace>/<bucket>.

This PR adds hf://buckets/... support to httpfs alongside the existing hf://datasets/... and hf://spaces/... handling.

Example usage

Verified locally against the public bucket cfahlgren1/duckdb-httpfs-fineweb-edu-5gb:

SELECT filename, token_count
FROM parquet_scan(
    'hf://buckets/cfahlgren1/duckdb-httpfs-fineweb-edu-5gb/data/**/*.parquet',
    filename = true
)
LIMIT 1;

I tested locally against the following buckets:

Public bucket:

• hf://buckets/cfahlgren1/duckdb-httpfs-fineweb-edu-5gb/data/**/*.parquet

Private bucket:

• hf://buckets/cfahlgren1/duckdb-httpfs-fineweb-edu-5gb-private/README.md
• verified with CREATE SECRET ... TYPE HUGGINGFACE, PROVIDER credential_chain

Changes

• accept buckets as a valid Hugging Face repo type
• treat buckets as non-revisioned resources
• build bucket file URLs without inserting a revision segment

[cfahlgren1]

A core contributor would need to create a small bucket fixture so we can add hf://buckets/... integration coverage in here, similar to the existing remote Hugging Face fixtures there under samansmink/duckdb_ci_tests for public datasets, private datasets, and spaces?

[lhoestq]

Yay ! let us know @carlopi if it looks good to you :)

[cfahlgren1]

@carlopi friendly bump if this makes sense to add 🤗

[nleroy917]

Very interested in this!!!

[lhoestq]

Gentle ping on this one :)

[davanstrien]

This sounds great!