feat: stream-based overloads to avoid byte[] allocations

[stevehansen]

Summary

Add Stream-based overloads for create, verify, and extract operations to avoid loading entire files into byte[], enabling support for large files with lower memory pressure.

Problem

Current API requires byte[] for all operations. For large files (hundreds of MB or more), this causes unnecessary memory allocations and prevents streaming scenarios. The CreateAsync(Stream, ...) overload exists but internally copies to byte[].

Proposed Solution

Create

• CreateAsync(Stream data, string fileName, Stream output, CancellationToken) — write the container directly to an output stream without materializing the full container in memory
• Hash the input stream incrementally instead of buffering

Verify

• VerifyAsync(Stream containerStream) — read and verify from a stream without loading the full container into byte[]
• Note: ZipArchive already accepts streams, so this is mostly API surface

Extract

• ExtractAsync(Stream containerStream, Stream output) — extract data file directly to an output stream

CLI

• The CLI already works with file paths, so it benefits automatically from stream-based internals

Considerations

• ZipArchive requires a seekable stream for read mode — may need to document this limitation or buffer selectively
• Hashing can be done incrementally via IncrementalHash or CryptoStream
• Backward compatible — existing byte[] overloads remain, new stream overloads are additive