feat: add burntpix benchmark and unify execution paths

[DaniPopes]

Add the BurntPix on-chain generative art contracts as a fixture-based benchmark (Cancun spec).

Unify the CLI and criterion benchmark execution so both bytecode-only and fixture benchmarks go through the same transact() flow. Bytecode benchmarks are wrapped into a synthetic fixture with a BENCH_CONTRACT and BENCH_CALLER. Benchmarks using stack_input are detected via is_runnable() and skipped for transaction-based execution.

Depends on #315.

[codspeed-hq]

Merging this PR will degrade performance by 99.98%

⚡ 2 improved benchmarks
❌ 15 regressed benchmarks
✅ 51 untouched benchmarks
🆕 18 new benchmarks
⏩ 1 skipped benchmark¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
🆕	airdrop/rt/jit	N/A	39.6 µs	N/A
❌	airdrop/rt/interpreter	6.6 µs	40.7 µs	-83.83%
🆕	fiat_token/rt/jit	N/A	35.2 µs	N/A
❌	factorial/rt/interpreter	266.1 µs	298 µs	-10.7%
❌	bswap64/rt/interpreter	10.3 µs	42.5 µs	-75.86%
🆕	factorial/rt/jit	N/A	53.8 µs	N/A
❌	eip4788/rt/interpreter	6.1 µs	38.9 µs	-84.29%
🆕	bswap64/rt/jit	N/A	39.3 µs	N/A
🆕	burntpix/rt/jit	N/A	87.1 ms	N/A
🆕	eip4788/rt/jit	N/A	38.5 µs	N/A
🆕	burntpix/rt/interpreter	N/A	280.3 ms	N/A
❌	fiat_token/rt/interpreter	2.9 µs	35.6 µs	-91.92%
🆕	bswap64_opt/rt/jit	N/A	38 µs	N/A
❌	erc20_transfer/rt/interpreter	10.2 µs	45,747.5 µs	-99.98%
🆕	counter/rt/jit	N/A	38.2 µs	N/A
❌	bswap64_opt/rt/interpreter	9.8 µs	42 µs	-76.66%
❌	counter/rt/interpreter	6.3 µs	39.7 µs	-84.03%
🆕	erc20_transfer/rt/jit	N/A	39 ms	N/A
🆕	eip2935/rt/jit	N/A	38 µs	N/A
❌	eip2935/rt/interpreter	5.8 µs	38.9 µs	-84.95%
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

---

_{Comparing dani/unified-bench (91ff075) with main (24772e2)}

Footnotes

1. 1 benchmark was skipped, so the baseline result was used instead. If it was deleted from the codebase, click here and archive it to remove it from the performance reports. ↩