ethexe-malachite: audit follow-up — test gaps and stale insert comment

[grishasobol]

Problem

Several small follow-ups from the May 2026 audit are tracked in design notes but not as code. Bundled here because each is too small to file separately:

1. chain_head == None branch in validate_block_above is not unit-tested. Commit 6d302a7a0 added an instant-abstain when local chain_head is unset, but no test exercises it.

2. TryFrom<RawProposedValue> bound check has no positive regression test. Commit 503a3d43d flipped silent From to fallible TryFrom, but the pinned-test that documented the bug was deleted along with the fix. A future refactor could re-introduce silent aliasing without breaking any test.

3. InjectedTxMempool::insert doc comment lines 326-340 still describes the old retain policy. Commit d52c62e01 changed purge_expired to drop unresolved ref-block txs (symmetric with seen), but the inline comment in insert was not updated.

Proposed solution

A single PR with three small additions:

1. validate_rejects_advance_when_chain_head_is_none() in service/src/externalities.rs tests module.
2. raw_proposed_value_out_of_range_round_is_rejected() in core/src/codec.rs tests module — assert TryFrom returns Err on round = i64::MIN.
3. Edit the doc comment in service/src/mempool.rs around the insert impl to reflect the post-commit-d52c62e01 drop policy.

Alternatives

File three separate issues — overkill for the scope of each fix.

Additional context

All three are audit follow-ups that surfaced while triaging issues already fixed (#7, #9, #10 in the audit tracker).

[AmIrRX0]

Hi! I would like to take this issue using Codex through Fixel as an AI-assisted workflow, with full human review before submission.

I will keep the change to the three audit follow-ups listed here, run the targeted ethexe-malachite tests plus the required repository checks, and disclose the assistance according to the project policy. Since this issue was opened in May, could you please confirm that all three items are still current and assign it to @AmIrRX0 if formal ownership is preferred? I will wait for that confirmation before opening a PR.