perf: cache packed proof nonces for cheaper header/block hash

[iho]

Summary

block.hash() / header.hash() ultimately hash the PoW Proof by bit-packing all nonces on every call. That packing is relatively expensive and runs very often during sync and validation.

This caches the packed form when a Proof is deserialized (the on-wire bytes we already read), so later hash/write uses that buffer instead of re-running pack_bits.

Closes #3372

Approach

• Add private packed_nonces: Option<Vec<u8>> on Proof (ignored by PartialEq / serde)
• Readable::read stores the packed bytes alongside extracted nonces
• Writeable::write prefers the cache; otherwise packs on demand (mining / constructed proofs)
• Proof::clear_packed_cache() for safe mutation of edge_bits / nonces
• Stratum share path clears cache after applying a solution
• Unit tests: round-trip cache, hash equivalence, mutation + clear

Locally constructed proofs (mining templates, tests, genesis) keep None and behave as before.

Test plan

• cargo test -p grin_core --lib pow::types
• cargo test -p grin_core --lib genesis
• cargo check -p grin_servers