Update Split and Splice docs

[tyler-french]

This PR better aligns the language of the REv2 API to describe how a client should expect to interact with the Split and Splice APIs.

Generally speaking, when designing a Remote Cache service, the server is not always primarily responsible for doing splitting and splicing blobs. In fact, the Split and Splice APIs are extremely helpful from a client's context to store and retrieve this manifest for how content defined chunking can compose a blob.

For example, if a client calls Splice that maps blob digest A to chunks A1 and A2, this instructs the server to store this information. Later, if a client that is not chunking-aware calls Read on A, the server can use this stored state to compose A from A1 and A2 stored in the CAS, and serve it to the client.

Similarly, if a user calls Split on blob B (which could be some Action Result), the server would respond with its stored manifest: B1 and B2. A chunking aware client can then skip downloading B1 if it's available locally from some other file's chunks, and download only B2 without ever needing to download the entirety of B.

[Copilot]

Pull request overview

This PR updates the documentation for the SplitBlob and SpliceBlob APIs in the Remote Execution v2 protocol to clarify their purpose and usage patterns. The changes emphasize that these APIs are primarily for storing and retrieving chunk composition metadata rather than performing the actual splitting and splicing operations on the server.

Key Changes:

• Clarified that SplitBlob retrieves stored information about how a blob is chunked rather than performing the split operation
• Updated SpliceBlob documentation to emphasize that clients tell the server how chunks compose a blob
• Expanded error conditions for SplitBlob to include cases where split information or chunks are missing

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[tyler-french]

@tjgq Do you have time to take another look here also? Thanks!