bench: add subcall benchmarks (1000-call variants)

[rakita]

Summary

• Add three subcall benchmark variants to revme bench for measuring CALL opcode performance
• subcall_1000_transfer_1wei: 1000 subcalls each transferring 1 wei (measures value transfer overhead)
• subcall_1000_same_account: 1000 subcalls with no value to same account (baseline subcall cost)
• subcall_1000_nested: 1000 subcalls where each target does another subcall (measures nested call frame overhead)

Test plan

• cargo build -p revme compiles successfully
• cargo run -p revme --release -- bench subcall runs all 3 variants (~157-299µs per iteration)

[codspeed-hq]

Merging this PR will improve performance by 3.72%

⚡ 2 improved benchmarks
✅ 171 untouched benchmarks
🆕 3 new benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	transfer	14.1 µs	13.6 µs	+3.72%
🆕	subcall_1000_transfer_1wei	N/A	1.2 ms	N/A
🆕	subcall_1000_nested	N/A	2.1 ms	N/A
🆕	subcall_1000_same_account	N/A	1.1 ms	N/A
⚡	transfer_finalize	17.5 µs	17 µs	+3.39%

---

_{Comparing subcall-bench (cb82522) with main (0b34496)}

[rakita]

Just a bench in revme, it does not touch crates at all