A way to specify offset/length when compressing and uncompressing

If I want to compress or uncompress only a view of my string/data, I don't see a way to do that without creating a copy. Do you think this can be added.

The issue body is the available starting point; inspect the Python bindings and their compression and decompression entry points to understand the current data-copy behavior. Define offset and length semantics for both operations, then add coverage showing that a view can be processed without creating a full copy.

索引モデルが issue の本文から書いたものです。