Import Telegram QA evidence RTT rows

[RomneyDa]

Summary

• Switch Telegram RTT workflows from the removed OpenClaw pnpm rtt wrapper to the package Telegram live QA lane.
• Make scripts/import-result.mjs consume OpenClaw qa-evidence.json directly for Telegram imports and continue writing the existing data/channels/telegram row shape used by the dashboard.
• Require aggregate Telegram RTT timing (result.timing.samples) before importing, so older/lagging harness output cannot silently record one-shot Telegram latency as release RTT data.
• Remove the old Telegram result.json importer path; historical rows stay in place, while future rows preserve aggregate timing from QA evidence.
• Keep workflow RTT env wiring minimal by relying on OpenClaw defaults where they match, while still allowing the main RTT workflow to override sample count for manual dispatch runs.

Pairs with openclaw/openclaw#92550. That OpenClaw PR provides the package Telegram live RTT controls and qa-evidence.json timing this repo imports. Older beta/release packages can still be the system under test because the checked-out OpenClaw harness owns the QA evidence shape.

Verification

• npm test
• npm run check
• git diff --check
• node --check scripts/*.mjs
• node scripts/validate.mjs
• node scripts/update-readme.mjs

Real behavior proof

Behavior addressed: openclaw-rtt no longer calls the removed OpenClaw pnpm rtt wrapper; Telegram main/release workflows import the standard QA evidence artifact instead.

Real environment tested: Local openclaw-rtt worktree on macOS with Node test/check commands.

Exact steps or command run after this patch: Ran the Node test suite, full repo check, whitespace check, script syntax check, data validation, and README generation.

Evidence after fix: npm test passed 39 tests; npm run check passed validation and README generation with a clean tree.

Observed result after fix: A synthetic Telegram qa-evidence.json import writes the same dashboard-critical row fields: package/run/channel metadata, canary RTT, mention RTT, avg/p50/p95/max, failed sample count, resources, and copied runs/telegram/<run>/result.json path. A synthetic non-aggregate qa-evidence.json is rejected before import, preventing old-harness one-shot evidence from polluting RTT rows.

What was not tested: No live Blacksmith/Telegram workflow was executed locally.