[WIP] perf: remove TrieUpdates::removed_nodes and StorageTrieUpdates::removed_nodes (attempt 2)

[kien-rise]

Replaces #13872. This PR is only slightly worse than as good as #13872, but the code changes are minimal.

Checklist

• E2E Benchmark
• Add benches/*
• Improve PR description

Motivation

Under high load (TPS > 50,000), MemoryOverlayStateProviderRef::trie_state is taking a considerable amount of time (over 200ms). One factor contributing to this is the TrieUpdates::extend_ref function. Optimizing the struct definition of TrieUpdates could help improve the performance of the extend_ref function.

Notable changes

pub struct TrieUpdates {
-    pub account_nodes: HashMap<Nibbles, BranchNodeCompact>,
-    pub removed_nodes: HashSet<Nibbles>,
+    pub changed_nodes: HashMap<Nibbles, Option<BranchNodeCompact>>,
     pub storage_tries: B256HashMap<StorageTrieUpdates>,
 }

pub struct StorageTrieUpdates {
     pub is_deleted: bool,
-    pub storage_nodes: HashMap<Nibbles, BranchNodeCompact>,
-    pub removed_nodes: HashSet<Nibbles>,
+    pub changed_nodes: HashMap<Nibbles, Option<BranchNodeCompact>>,
 }

Benchmarks

192.155.101.220/slave/785bc168-9976731c-erc20-30600/realtime-block.stdout.log
{"tps": 30601.0, "gps": 1056921850.8367347, "is_chain_lagged": false, "chain_lag_distance": 0}
192.155.101.220/slave/785bc168-9976731c-erc20-31000/realtime-block.stdout.log
{"tps": 31001.0, "gps": 1070736094.1406412, "is_chain_lagged": false, "chain_lag_distance": 0}
192.155.101.220/slave/785bc168-9976731c-erc20-32000/realtime-block.stdout.log
{"tps": 32001.0, "gps": 1105272921.4311633, "is_chain_lagged": false, "chain_lag_distance": 1}
192.155.101.220/slave/785bc168-9976731c-erc20-33000/realtime-block.stdout.log
{"tps": 33001.0, "gps": 1139812889.5115511, "is_chain_lagged": false, "chain_lag_distance": 1}
192.155.101.220/slave/785bc168-9976731c-erc20-34000/realtime-block.stdout.log
{"tps": 34001.0, "gps": 1174346187.1292517, "is_chain_lagged": true, "chain_lag_distance": 4}
192.155.101.220/slave/785bc168-9976731c-erc20-36000/realtime-block.stdout.log
{"tps": 36001.0, "gps": 1243424508.7788463, "is_chain_lagged": true, "chain_lag_distance": 20}
192.155.101.220/slave/785bc168-9976731c-erc20-40000/realtime-block.stdout.log
{"tps": 40001.0, "gps": 1381586516.224, "is_chain_lagged": true, "chain_lag_distance": 61}
192.155.101.220/slave/785bc168-9976731c-raw-transfer-54600/realtime-block.stdout.log
{"tps": 54601.0, "gps": 1146644512.2513661, "is_chain_lagged": false, "chain_lag_distance": 1}
192.155.101.220/slave/785bc168-9976731c-raw-transfer-55000/realtime-block.stdout.log
{"tps": 55001.0, "gps": 1155044519.4926472, "is_chain_lagged": false, "chain_lag_distance": 0}
192.155.101.220/slave/785bc168-9976731c-raw-transfer-56000/realtime-block.stdout.log
{"tps": 56001.0, "gps": 1176044515.2560747, "is_chain_lagged": false, "chain_lag_distance": 1}
192.155.101.220/slave/785bc168-9976731c-raw-transfer-57000/realtime-block.stdout.log
{"tps": 57001.0, "gps": 1197044511.5361216, "is_chain_lagged": false, "chain_lag_distance": 0}
192.155.101.220/slave/785bc168-9976731c-raw-transfer-58000/realtime-block.stdout.log
{"tps": 58001.0, "gps": 1218044525.3953488, "is_chain_lagged": false, "chain_lag_distance": 1}
192.155.101.220/slave/785bc168-9976731c-uniswap-6330/realtime-block.stdout.log
{"tps": 6331.0, "gps": 882450529.3786408, "is_chain_lagged": false, "chain_lag_distance": 0}
192.155.101.220/slave/785bc168-9976731c-uniswap-6400/realtime-block.stdout.log
{"tps": 6401.0, "gps": 892212483.2733675, "is_chain_lagged": false, "chain_lag_distance": 0}
192.155.101.220/slave/785bc168-9976731c-uniswap-6480/realtime-block.stdout.log
{"tps": 6481.0, "gps": 903368570.772521, "is_chain_lagged": false, "chain_lag_distance": 0}
192.155.101.220/slave/785bc168-9976731c-uniswap-6580/realtime-block.stdout.log
{"tps": 6581.0, "gps": 917296337.1077226, "is_chain_lagged": true, "chain_lag_distance": 4}