Refactor sector proof to use cached subtree roots

[n8mgr]

Refactors sector proofs to use cached subtree roots to reduce disk IO when building a proof for part of a sector. (SiaFoundation/hostd#876, SiaFoundation/coreutils#364)

(Yes, I copied sectorAccumulator in but it's probably time we remove the RHP2 and RHP3 packages in a follow up)

@lukechampine feel free to take this over if you are unhappy with it.

[Copilot]

Pull request overview

This PR refactors sector proof generation to use cached subtree roots at a fixed 64-leaf level, reducing disk I/O operations by allowing hosts to build proofs without reading the entire sector data. This optimization is particularly beneficial for reducing the amount of data read from disk during proof generation.

Key Changes:

• Introduces a specialized sectorAccumulator for efficient Merkle tree computation
• Adds CachedSectorSubtrees to compute and cache 64-leaf subtree roots (4 KiB chunks)
• Refactors BuildSectorProof to construct proofs using cached subtrees and minimal segment data

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 5 comments.

File	Description
rhp/v4/merkle.go	Adds caching infrastructure with sectorAccumulator, CachedSectorSubtrees, and refactored BuildSectorProof to use cached 64-leaf subtree roots
rhp/v4/merkle_test.go	Adds comprehensive test coverage for BuildSectorProof with various range scenarios including edge cases and randomized inputs

---

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

[lukechampine]

LGTM. I agree that it's time to absorb the RHP2 and RHP3 packages into RHPv4.

[peterjan]

This'll spin up 1024 goroutines per sector right? Isn't that an excessive amount? Obviously it would come at the cost of a little more complexity but perhaps we could batch these and use runtime.GOMAXPROCS or use a pool or something?