perf: make Chain use DeferredTrieData

[Rjected]

This makes Chain use the DeferredTrieData instead of HashedPostStateSorted and TrieUpdatesSorted.

Key Changes

• Deferred Trie Data: The Chain type now stores BTreeMap<BlockNumber, DeferredTrieData> instead of separate maps for trie updates and hashed state
• Non-blocking Construction: Chain can be constructed without blocking on expensive trie computations
• Lazy Access: Trie data is computed in background tasks and materialized only when needed

Accessing Trie Data

The trie data is still fully accessible, but access may block if background computation is in progress:

• chain.trie_data() / chain.trie_data_at(num) - Returns DeferredTrieData handles (non-blocking)
• chain.trie_updates() / chain.trie_updates_at(num) - Blocks until computation completes, then returns Arc<TrieUpdatesSorted>
• chain.hashed_state() / chain.hashed_state_at(num) - Blocks until computation completes, then returns Arc<HashedPostStateSorted>

For best performance, prefer using trie_data() and calling wait_cloned() explicitly when you need the computed data.

Crate Structure

A new std-only crate reth-chain is added which contains:

• Chain type
• DeferredTrieData and ComputedTrieData
• Bincode-compatible serde implementations for WAL persistence

Backwards Compatibility

WAL files created with older versions (without trie data) will deserialize correctly - missing fields default to empty maps.

[mattsse]

makes sense

[mattsse]

imo this is fine

[codspeed-hq]

Merging this PR will not alter performance

✅ 118 untouched benchmarks
⏩ 7 skipped benchmarks¹

---

_{Comparing dan/chain-deferred-1 (b4fd9dd) with main (b96a308)}

Footnotes

1. 7 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩