Replace BytesTrieSet with HashSet

[siladu]

Related to #9165
(Replaces #9166)

This PR effectively reverts #6069 by wholesale removing BytesTrieSet and going back to using HashSet as was the case with this PR: #5537

Benchmarks show improvements across CREATE, CREATE2, SELFDESTRUCT and jumpdest analysis and no regressions including *CALL.

---

No mainnet improvement/regression detected, newPayload response time p50:

---

Performed benchmarking using evm block-test and the eest benchmark compute suite as of ethereum/execution-specs@2c83b84

Ran https://github.com/ethereum/execution-specs/tree/forks/amsterdam/tests/benchmark/compute/instruction
and https://github.com/ethereum/execution-specs/tree/forks/amsterdam/tests/benchmark/compute/scenario
ignoring the precompile suite.

Instruction Benchmark: Control vs replace-BytesTrieSet

  Summary
  ┌────────────────────────────┬────────────────────────────┬────────────────────────────┬────────────────────┐
  │         Comparison         │          Control           │            Test            │     Difference     │
  ├────────────────────────────┼────────────────────────────┼────────────────────────────┼────────────────────┤
  │ All runs (7 ctrl, 3 test)  │ 463,541 ms @ 165.69 MGas/s │ 454,614 ms @ 168.77 MGas/s │ -8,927 ms (-1.93%) │
  ├────────────────────────────┼────────────────────────────┼────────────────────────────┼────────────────────┤
  │ repeat2 only (2 vs 3 runs) │ 461,082 ms @ 166.11 MGas/s │ 454,614 ms @ 168.77 MGas/s │ -6,468 ms (-1.40%) │
  └────────────────────────────┴────────────────────────────┴────────────────────────────┴────────────────────┘
  Top 10 Improvements
  ┌──────┬──────────────────┬─────────────────────────────────────────────────┐
  │ Rank │      Change      │                      Test                       │
  ├──────┼──────────────────┼─────────────────────────────────────────────────┤
  │ 1    │ -47.7% (-219 ms) │ test_create[0 bytes with value-CREATE2]         │
  ├──────┼──────────────────┼─────────────────────────────────────────────────┤
  │ 2    │ -46.3% (-206 ms) │ test_create[0 bytes without value-CREATE2]      │
  ├──────┼──────────────────┼─────────────────────────────────────────────────┤
  │ 3    │ -45.6% (-200 ms) │ test_create[0 bytes with value-CREATE]          │
  ├──────┼──────────────────┼─────────────────────────────────────────────────┤
  │ 4    │ -44.4% (-283 ms) │ test_selfdestruct_initcode[value_bearing_True]  │
  ├──────┼──────────────────┼─────────────────────────────────────────────────┤
  │ 5    │ -44.4% (-191 ms) │ test_create[0 bytes without value-CREATE]       │
  ├──────┼──────────────────┼─────────────────────────────────────────────────┤
  │ 6    │ -44.3% (-288 ms) │ test_selfdestruct_initcode[value_bearing_False] │
  ├──────┼──────────────────┼─────────────────────────────────────────────────┤
  │ 7    │ -38.8% (-2 ms)   │ test_create[max code size-CREATE2]              │
  ├──────┼──────────────────┼─────────────────────────────────────────────────┤
  │ 8    │ -35.8% (-420 ms) │ test_selfdestruct_created[value_bearing_True]   │
  └──────┴──────────────────┴─────────────────────────────────────────────────┘
  Top Regressions (small absolute values)
  ┌──────┬────────────────────┬────────────────────────────────────────────────┐
  │ Rank │       Change       │                      Test                      │
  ├──────┼────────────────────┼────────────────────────────────────────────────┤
  │ 1    │ +103.2% (+1.54 ms) │ test_log[zeros_topic-1_MiB_zeros_data-log1]    │
  ├──────┼────────────────────┼────────────────────────────────────────────────┤
  │ 2    │ +99.0% (+1.49 ms)  │ test_log[zeros_topic-1_MiB_non_zero_data-log2] │
  ├──────┼────────────────────┼────────────────────────────────────────────────┤
  │ 3    │ +84.7% (+1.43 ms)  │ test_log[zeros_topic-1_MiB_zeros_data-log4]    │
  └──────┴────────────────────┴────────────────────────────────────────────────┘
  Note: The regressions are all sub-2ms tests - the high percentages are misleading. The improvements are in the 200-420ms range, far outweighing the tiny regressions.

  Conclusion

  replace-BytesTrieSet is ~1.4-1.9% faster overall, with significant improvements in CREATE, CREATE2, and SELFDESTRUCT operations. The log test "regressions" are noise in sub-2ms tests.

Scenario Benchmark Results: Control vs replace-BytesTrieSet

Methodology: 10 iterations each, using evmtool block-test with execution-focused tests

Summary

Metric	Control	Test	Difference
Total Time	6232.49 ms	5545.00 ms	-687.49 ms (-11.0%)
Throughput	364.83 MGas/s	410.60 MGas/s	+45.77 MGas/s
Tests	24	24	-
Iterations	10	10	-

Result: Test is 11.0% faster (687 ms saved)

Top Improvements

Test	Control	Test	Change
test_jumpdest_analysis[benchmark-gas-value_100M-fork_Prag...	452.48 ms	336.74 ms	-25.6%
test_jumpdest_analysis[benchmark-gas-value_100M-fork_Prag...	469.16 ms	352.35 ms	-24.9%
test_jumpdest_analysis[benchmark-gas-value_100M-fork_Prag...	453.07 ms	346.33 ms	-23.6%
test_jumpdest_analysis[benchmark-gas-value_100M-fork_Prag...	497.59 ms	383.95 ms	-22.8%
test_jumpdest_analysis[benchmark-gas-value_100M-fork_Prag...	477.13 ms	375.30 ms	-21.3%
test_jumpdest_analysis[benchmark-gas-value_100M-fork_Prag...	503.35 ms	404.36 ms	-19.7%
test_block_full_of_ether_transfers[benchmark-gas-value_10...	104.38 ms	99.45 ms	-4.7%
test_block_full_of_ether_transfers[benchmark-gas-value_10...	354.88 ms	339.69 ms	-4.3%
test_block_full_of_ether_transfers[benchmark-gas-value_10...	210.20 ms	202.09 ms	-3.9%
test_auth_transaction[benchmark-gas-value_100M-fork_Pragu...	327.28 ms	316.97 ms	-3.2%

Potential Regressions

Test	Control	Test	Change
test_block_full_of_ether_transfers[benchmark-gas-value_10...	211.63 ms	222.87 ms	+5.3%
test_block_full_of_ether_transfers[benchmark-gas-value_10...	111.42 ms	114.89 ms	+3.1%
test_block_full_access_list_and_data[benchmark-gas-value_...	53.18 ms	54.72 ms	+2.9%
test_block_full_of_ether_transfers[benchmark-gas-value_10...	113.49 ms	115.54 ms	+1.8%
test_block_full_data[benchmark-gas-value_100M-fork_Prague...	51.18 ms	52.02 ms	+1.6%

Note: Small regressions (<5%) are within measurement variance (CV typically 5-15%)

All Tests (sorted by improvement)

#	Test	Control (ms)	Test (ms)	Diff	Ctrl CV	Test CV
1	test_jumpdest_analysis[benchmark-gas-value_100M...	452.48	336.74	-25.6%	1.4%	8.7%
2	test_jumpdest_analysis[benchmark-gas-value_100M...	469.16	352.35	-24.9%	0.8%	7.2%
3	test_jumpdest_analysis[benchmark-gas-value_100M...	453.07	346.33	-23.6%	1.6%	7.8%
4	test_jumpdest_analysis[benchmark-gas-value_100M...	497.59	383.95	-22.8%	1.1%	6.2%
5	test_jumpdest_analysis[benchmark-gas-value_100M...	477.13	375.30	-21.3%	1.0%	7.7%
6	test_jumpdest_analysis[benchmark-gas-value_100M...	503.35	404.36	-19.7%	1.0%	6.5%
7	test_block_full_of_ether_transfers[benchmark-ga...	104.38	99.45	-4.7%	16.2%	13.0%
8	test_block_full_of_ether_transfers[benchmark-ga...	354.88	339.69	-4.3%	8.3%	5.1%
9	test_block_full_of_ether_transfers[benchmark-ga...	210.20	202.09	-3.9%	17.5%	18.6%
10	test_auth_transaction[benchmark-gas-value_100M-...	327.28	316.97	-3.2%	3.9%	0.9%
11	test_block_full_of_ether_transfers[benchmark-ga...	111.68	108.33	-3.0%	13.5%	14.3%
12	test_block_full_data[benchmark-gas-value_100M-f...	15.12	14.67	-2.9%	9.7%	6.0%
13	test_block_full_of_ether_transfers[benchmark-ga...	228.92	223.80	-2.2%	14.2%	14.7%
14	test_block_full_of_ether_transfers[benchmark-ga...	338.85	332.40	-1.9%	11.2%	9.2%
15	test_block_full_of_ether_transfers[benchmark-ga...	232.31	228.66	-1.6%	11.8%	11.0%
16	test_empty_block[benchmark-gas-value_100M-fork_...	0.97	0.96	-0.5%	3.3%	3.3%
17	test_auth_transaction[benchmark-gas-value_100M-...	284.67	284.28	-0.1%	3.9%	4.3%
18	test_auth_transaction[benchmark-gas-value_100M-...	294.89	295.53	+0.2%	1.5%	4.4%
19	test_auth_transaction[benchmark-gas-value_100M-...	334.67	339.09	+1.3%	1.9%	2.6%
20	test_block_full_data[benchmark-gas-value_100M-f...	51.18	52.02	+1.6%	3.1%	2.2%
21	test_block_full_of_ether_transfers[benchmark-ga...	113.49	115.54	+1.8%	6.0%	11.3%
22	test_block_full_access_list_and_data[benchmark-...	53.18	54.72	+2.9%	8.0%	11.5%
23	test_block_full_of_ether_transfers[benchmark-ga...	111.42	114.89	+3.1%	7.6%	9.6%
24	test_block_full_of_ether_transfers[benchmark-ga...	211.63	222.87	+5.3%	9.0%	9.2%

[lu-pinto]

LGTM

[ahamlat]

🚢