perf(re-execute): relax executor reset thresholds

[emmajam]

Increase state cache size limit from 1M to 5M entries and executor lifetime from 120s to 600s. Preserves hot state caches longer during re-execute, reducing redundant DB lookups for frequently accessed accounts and storage slots.

Benchmark (dev-emma, 3 runs, blocks 21M–21.001M)

	Baseline (main)	Feature
Run 1	0.478 Ggas/s	0.520 Ggas/s
Run 2	0.486 Ggas/s	0.520 Ggas/s
Run 3	0.480 Ggas/s	0.512 Ggas/s
Mean	0.481 Ggas/s	0.517 Ggas/s

+7.5% ±1.5% (sig=good) — statistically significant at 95% CI via Welch's t-test.

Changes

• crates/cli/commands/src/re_execute.rs: size_hint threshold 1M → 5M, executor_lifetime 120s → 600s

[emmajam]

50K block benchmark (dev-emma, 3 runs, blocks 21M–21.05M)

	Baseline (main)	Feature
Run 1	2.39 Ggas/s	2.73 Ggas/s
Run 2	2.49 Ggas/s	2.80 Ggas/s
Run 3	2.50 Ggas/s	2.82 Ggas/s
Mean	2.46 Ggas/s	2.78 Ggas/s

+13.1% ±3.5% (sig=good) — stronger effect at scale since more executor resets are avoided over 50K blocks.

[mattsse]

seems fine

[mattsse]

this exceeds the 5min threshold we usually have but in this case it doesnt really matter and I believe we'll always encounter the size hint first

[emmajam]

just benching the change on tempo also

[emmajam]

Tempo benchmark — no improvement

Tested the same change on Tempo chain data (21 GB, blocks 14M–15M, dev-emma).

1K blocks (15M–15.001M):

	Baseline	Feature
Mean	2.26 Ggas/s	2.31 Ggas/s

1M blocks (14M–15M):

	Baseline	Feature
Run 1	86.62 Ggas/s	87.80 Ggas/s
Run 2	88.28 Ggas/s	87.54 Ggas/s
Run 3	88.56 Ggas/s	87.61 Ggas/s
Mean	87.82 Ggas/s	87.65 Ggas/s

No meaningful difference on Tempo. The chain is much smaller (21 GB vs 2.5 TB for Ethereum mainnet) so state fits in cache easily and executor resets are essentially free — the bottleneck is elsewhere.