persisted_trie_updates data lost on restart prevents node from handling reorgs

[klkvr]

Describe the feature

If node is restarted, then we are losing persisted_trie_updates data in TreeState. This results in executed_block_by_hash always returning None for canonical blocks

reth/crates/engine/tree/src/tree/mod.rs

Lines 1593 to 1595 in f93d307

	let Some((_, updates)) = self.state.tree_state.persisted_trie_updates.get(&hash) else {
	return Ok(None)
	};

This results in node not being able to handle reorgs by ending up here

reth/crates/engine/tree/src/tree/mod.rs

Lines 943 to 945 in f93d307

	// This shouldn't happen as we're walking back the canonical chain
	warn!(target: "engine::tree", current_hash=?old_hash, "Canonical block not found in TreeState");
	return Ok(None);

Solution for this is to ignore cache misses in persisted_trie_updates for canonical blocks because we don't need TrieUpdates for reorg of the canonical chain, and for sidechains they are always present because those live fully in memory

Additional context

No response