Add BlobStore batch operations

[captjt]

Summary

This is the integrity SDK companion change for GOV-342. It adds batch blob-store operations at the trait boundary so callers that register or resolve many lineage blobs can avoid one fully serial blob-store request per CID. The default implementation runs existing single-object operations with bounded async concurrency, while leaving room for provider-specific overrides where a backend supports a native batch API.

The PR also exposes the new batch operations through the FFI layer used by the Python SDK, so downstream services can call the SDK batch path without waiting on the CLI/core-rs path.

Problem

Lineage registration and resolution can involve many blobs, especially when an Iroh collection references multiple file blobs. The existing BlobStore API only exposed exists, get, and put, so every caller had to issue individual blob operations. In service paths this adds avoidable latency from serial network round trips and prevents storage backends from optimizing at their own layer.

S3 existence checks were also doing paginated object listing to determine whether a single CID existed, which is much more expensive than a direct object metadata check.

Changes

• Added BlobPut, BlobPutResult, BlobGetResult, and BlobExistsResult types.
• Added BlobStore::exists_many, BlobStore::get_many, and BlobStore::put_many default methods.
• Added a bounded default concurrency limit for batch methods so all current providers immediately benefit without provider rewrites.
• Updated S3 exists to use HeadObject against the exact object key instead of paginated ListObjectsV2.
• Added FFI batch APIs:
  • ig_blob_store_exists_many
  • ig_blob_store_get_many
  • ig_blob_store_put_many
  • matching result-free helpers
• Updated manifest blob resolution so Iroh collection file blobs are fetched through get_many.
• Added tests for default batch method ordering/missing-blob behavior and FFI local filesystem batch roundtrips.

Validation

• cargo fmt --all
• cargo check -p integrity-blob --features blob-all
• cargo test -p integrity-blob
• cargo test -p integrity-ffi
• cargo test -p integrity-lineage-models
• cargo check -p integrity --no-default-features --features blob-all,lineage
• git diff --check

Follow-up

After this merges, we should mint a new integrity SDK version and update the integrity-service/Python SDK dependency to consume these FFI batch APIs. Native provider batch overrides can be added later where a backend supports them directly.

[TJbrunk]

The python sdk doesn't use any FFI bindings from this repo. It imports the rust code directly and creates it own bindings during compilation. Don't need to delete them, just an FYI.