chore(crypto): adopt NewKeccakState wrapper + global hasher pool

[colinkim]

Summary

• Adopt upstream go-ethereum's keccak-related optimizations in a targeted, low-risk way: the crypto.NewKeccakState() wrapper (from #29362) and the global hasher pool (from #31769).
• Stay behaviour-preserving — no public API signature, hash output, or thread-safety changes. Only the KeccakState allocation pattern changes.
• Apply a subset of each upstream PR: the NewKeccakState() wrapper from #29362 and the global hasherPool in crypto/crypto.go from #31769, because their current shape diverges from upstream enough that they cannot be applied directly.

Behaviour and compatibility

• Hash outputs: unchanged.
• Public API signatures: unchanged.
• Thread safety: unchanged. sync.Pool is goroutine-safe; hashers are not shared mid-operation across goroutines.
• golang.org/x/crypto/sha3 imports removed from files where no other sha3 symbol is used.

Test plan

• make gstable — build succeeds
• go vet ./... — no new warnings
• go test ./crypto/...
• go test ./core/rawdb/... ./core/types/... ./trie/... ./eth/protocols/snap/...
• go test -bench=BenchmarkKeccak -benchmem ./crypto/... — confirm allocs/op drops for Keccak256 / Keccak256Hash relative to the pre-PR baseline

References

• Upstream PR: core/rawdb, trie: improve db APIs for accessing trie nodes (commit 9f96e07c1)
• Upstream PR: all: create global hasher pool (commit 0eb2eeea9)