Tiered test suite + ClawHub publication readiness

[milstan]

Summary

• Tests (57 passing, ~2s, no network). Three tiers: contract (manifest ↔ registered-tool parity — catches drift with a named diff), unit (error mapping table, caching, tool branches), and opt-in live smoke (gated by LEADBAY_TEST_TOKEN). Harness in test/harness.ts isolates vi.mock('node:https') behind a single predicate-match helper with opinionated unmatched-request errors.
• Bug fix in src/client.ts. requestVoid only mapped 401, breaking the paid→org 404 fallback in enrich-contacts.ts. Extracted a shared mapErrorResponse helper; both request<T> and requestVoid now produce the same error codes.
• Manifest sync. openclaw.plugin.json now declares all 11 tools (adds leadbay_get_lead_activities and leadbay_get_taste_profile), version, and a sensitive token uiHint.
• Publication readiness per https://docs.openclaw.ai/plugins/sdk-setup: renames npm package to @leadbay/openclaw-leadclaw, adds openclaw.compat + openclaw.build blocks, files whitelist, prepublishOnly, CHANGELOG.md, and README Development + Publishing sections.

Test plan

• npm test — 57/57 green (contract + unit + sanity), no network
• npm run test:coverage — client.ts at 95%
• npm run test:smoke without token — cleanly skips with SMOKE_SKIPPED message
• npm run build — emits dist/ unchanged
• npm publish --dry-run — 33 files, 1.3 MB, test/ excluded
• Negative verification — flip 402 mapping → named row fails; remove tool from manifest → contract test prints exact diff
• LEADBAY_TEST_TOKEN smoke run against dedicated test tenant (CI / maintainer)
• Confirm 2026.3.24-beta.2 compat strings match current OpenClaw stable before publishing

🤖 Generated with Claude Code