fix(inspector): ignore stale SELFDESTRUCT journal entries

[0xalpharush]

Summary

• record the journal index immediately before each interpreter instruction
• derive SELFDESTRUCT callback metadata only from entries produced by the terminating instruction
• add a Cancun regression test for SELFDESTRUCT-to-self after a transaction value transfer

This reapplies the standalone fix from #3797 directly onto main, per the maintainer feedback in that PR.

Root cause

Under EIP-6780, SELFDESTRUCT-to-self can be a state no-op and append no destruction entry. The inspector previously read journal.last(), which could reuse an unrelated earlier balance transfer or destruction and emit a false SELFDESTRUCT callback with stale metadata.

Execution, state, and receipts are unaffected; this corrects inspector/tracer callbacks.

Validation

• cargo test -p revm-inspector cancun_selfdestruct_to_self_does_not_reuse_prior_journal_entry
• cargo test -p revm-inspector --lib
• cargo check -p revm-inspector --all-features
• cargo check -p revm-inspector --no-default-features
• cargo clippy -p revm-inspector --all-targets --all-features -- -D warnings
• cargo fmt --all --check

[codspeed-hq]

Merging this PR will not alter performance

✅ 181 untouched benchmarks
⏩ 1 skipped benchmark¹

---

_{Comparing 0xalpharush:agent/selfdestruct-inspector-journal-main (dee35be) with main (5199506)}

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. ↩