perf: stop scanning stale session markers (#153)

[dajiaohuang]

Summary

Address the session/history scan portion of issue #153. State restoration and context-marker lookup now walk from the newest entry backwards and stop at the first relevant marker, avoiding traversal of stale history while preserving the existing latest-marker semantics. A compatibility check covers the short-circuit behavior.

This is a focused first increment for #153; it does not claim to implement the issue's broader evaluation parallelism, context-size, or persistent-index proposals.

Authorship and provenance — select exactly one

• Human-authored
• Autonomous agent-authored
• Hybrid

Agent/tool and model/version: Codex (model/version not exposed in this session).

Agent contribution: Reviewed the performance analysis, selected the smallest behavior-preserving scan optimization, implemented it, added a short-circuit check, and ran the checks below.

Human verification: The submitting user authorized direct submission and reviewed the requested scope. The complete diff was reviewed by the agent; no independent human test execution is claimed.

Known limitations or uncertain results: No provider-backed benchmark was run, so this PR makes no quantified latency claim. The remaining performance proposals require separate acceptance criteria and measurement.

Labels

Target label: Target:Integrations

Author label: Author:AI

Workflow labels: enhancement

Safety and side effects

• The change does not access or expose secrets, private files, or unrelated user/repository data.
• Scripts, hooks, workflows, and evals are bounded and do not create surprising or irreversible side effects.
• No destructive, privileged, production, externally visible, or persistent action occurs without explicit user intent and appropriate safeguards.
• Network access, third-party code, permissions, and provider costs are minimized and documented.
• Prompt text, examples, and fixtures contain no hidden instructions that weaken safety or expand agent authority.

Side effects, permissions, network access, and cost: None. The changes alter only traversal order and stop once the newest relevant state marker is found; output semantics remain unchanged.

Compatibility

• This is not a breaking change.
• This is a breaking change; it was discussed, and migration/deprecation documentation is included below.
• Canonical and mirrored skill files are synchronized when applicable.
• Relevant platform manifests and installation documentation were reviewed.

Migration or rollback notes: None. If no relevant marker exists, both functions retain their previous default state.

Verification

• python -m unittest discover -s tests -v — 39 passed; 4 existing Windows eval-runner tests fail with FileNotFoundError: [WinError 2], reproduced on upstream main.
• python scripts/run_evals.py validate — passed.
• bun scripts/check_context_compat.ts — passed.
• git diff --check — passed.

Behavior evals: Not run; this is an internal traversal optimization with no response-rule change.

Final accountability

• I reviewed the complete diff, removed unrelated generated changes, and take responsibility for the submitted content.
• All failed, skipped, or unrun checks are disclosed above.