chore(l2): refactor get_embedded_node and remove hash cloning

[xqft]

Motivation

Optimizes trie decoding in zkVM stateless execution by adding NodeHash::finalize_mut and NodeRef::compute_hash_finalized to calculate a node's finalized hash, store it and return a reference to it, instead of cloning the hash or even hashing each time if it's an inline node.

Also refactors the get_embedded_node function into NodeRef::resolve_subtrie to remove the nested function and simplify the code.

Flamegraphs
After all other trie zkVM optimizations, from_nodes is one of the most expensive calls in block execution. This PR reduces Trie::from_nodes from 131k to 101k, 22%.

before:

after:

[github-actions]

Lines of code report

Total lines added: 60
Total lines removed: 35
Total lines changed: 95

Detailed view

+----------------------------------------+-------+------+
| File                                   | Lines | Diff |
+----------------------------------------+-------+------+
| ethrex/crates/common/trie/node.rs      | 228   | +45  |
+----------------------------------------+-------+------+
| ethrex/crates/common/trie/node_hash.rs | 119   | +15  |
+----------------------------------------+-------+------+
| ethrex/crates/common/trie/trie.rs      | 917   | -35  |
+----------------------------------------+-------+------+

[github-actions]

Benchmark for ad8c4f4

Click to view benchmark

Test	Base	PR	%
Trie/cita-trie insert 10k	33.2±1.16ms	32.6±0.56ms	-1.81%
Trie/cita-trie insert 1k	3.3±0.09ms	3.5±0.27ms	+6.06%
Trie/ethrex-trie insert 10k	61.9±1.55ms	62.0±1.32ms	+0.16%
Trie/ethrex-trie insert 1k	8.0±0.16ms	7.9±0.16ms	-1.25%

[Copilot]

Pull Request Overview

This PR refactors trie decoding to eliminate hash cloning and simplify the codebase. The changes optimize stateless execution in zkVM by introducing new methods for efficient hash computation and storage, resulting in a 22% performance improvement in Trie::from_nodes.

Key changes:

• Added NodeHash::finalize_mut to compute and store finalized hashes in-place, avoiding repeated hashing and clones
• Refactored the nested get_embedded_node function into NodeRef::resolve_subtrie for better code organization
• Introduced NodeRef::compute_hash_finalized to obtain hash references without cloning

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
crates/common/trie/trie.rs	Simplified Trie::from_nodes by replacing nested function with resolve_subtrie method
crates/common/trie/node_hash.rs	Added finalize_mut method to compute and cache finalized hash in-place
crates/common/trie/node.rs	Added compute_hash_finalized and resolve_subtrie methods to support refactored trie construction

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[github-actions]

Benchmark for 4bcaa8d

Click to view benchmark

Test	Base	PR	%
Trie/cita-trie insert 10k	41.5±2.54ms	36.6±1.59ms	-11.81%
Trie/cita-trie insert 1k	3.5±0.01ms	3.6±0.15ms	+2.86%
Trie/ethrex-trie insert 10k	52.3±1.99ms	55.0±2.31ms	+5.16%
Trie/ethrex-trie insert 1k	6.4±0.04ms	6.4±0.05ms	0.00%

[github-actions]

Benchmark for b5b710f

Click to view benchmark

Test	Base	PR	%
Trie/cita-trie insert 10k	34.8±0.58ms	35.0±0.83ms	+0.57%
Trie/cita-trie insert 1k	3.5±0.02ms	3.5±0.05ms	0.00%
Trie/ethrex-trie insert 10k	48.7±1.08ms	48.8±1.35ms	+0.21%
Trie/ethrex-trie insert 1k	6.4±0.13ms	6.3±0.03ms	-1.56%

[MegaRedHand]

This function leaves the tree in an inconsistent state, right? If any function expects compute_hash to return the node's MPT hash, they'll receive an incorrect result. We should document that in a big warning message to avoid weird errors going forward.