feat: mock atomizer and work-performer services with contract tests (phase 11)

- services/atomizer: standalone Go module implementing §5.1 — Anthropic
  OpenAI-compatible chat completions by default, native /v1/messages when
  LLM_API_STYLE=anthropic, strict-JSON prompting, defensive fence-stripping
  parse with one retry, deterministic equal-split fallback without an API
  key, exact coefficient-sum normalization, bearer auth
- services/work-performer: Node 20 http server implementing §5.2 — single
  concurrency, /work/{jobId}/TASK.md preparation, attachment downloads,
  optional shallow git clone, claude CLI execution with JSON output,
  simulated success when the CLI is unavailable (offline demo), artifact
  endpoint, idempotent HMAC-signed callbacks with retry, best-effort cancel
- compose profile 'mocks': separate builds/ports/tokens, healthchecks,
  ${HOME}/.claude(.json) mounted read-only into the performer
- contract tests (go test -tags=contract) for both services; Makefile
  test-contract target; verified live incl. a real Claude Code job run

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

PROGRESS.md

@@ -11,3 +11,4 @@
 - Phase 8 (bounty board + lifecycle): developer board (pool-scoped, filters/search/sort, stale age badges, hide-competing-claims setting, saved filters), claim/withdraw/decline/approve, assign-to-AI via §5.2 client + HMAC-verified idempotent callback w/ artifact ingestion into GridFS, start/submit/abandon/unassign, comments (sanitized, @mentions→notifications), time log, review queue + per-AC checklist on timeline, immutable bountyAwards on approve (humans only), notifications center + bell + WS toasts, my-tasks kanban, task detail page, pool management UI, server-side HTML sanitizer w/ XSS vector tests — full lifecycle integration tests green.
 - Phase 9 (messaging): conversations (dm dedupe/group/project), sanitized rich-text messages, file/image attachments (generic POST /api/v1/files w/ rate limit), per-conversation unread counts + mark-read, chat-file access limited to participants, mentions→notifications, typing indicator over WS inbound relay, two-pane messages UI (contenteditable composer, formatting toolbar, drag&drop upload, inline previews + lightbox, live delivery, polling fallback), user directory search — integration tests green.
 - Phase 10 (metrics): aggregation pipelines over bountyAwards + timelines (totals, weekly $dateTrunc buckets, per-developer/per-customer, approval rate, time logged, assign→approve lead time, atomization lead time, open board depth), CSV export, leaderboard w/ opt-out (profile setting), dependency-free SVG line+bar charts, developer/consultant/admin dashboards — integration tests green.
+- Phase 11 (mock services): services/atomizer (own Go module, ports 8090, bearer auth, OpenAI-compatible chat completions + native /v1/messages via LLM_API_STYLE, strict-JSON prompt, fence-stripping defensive parse, one retry, deterministic equal-split fallback, exact sum normalization); services/work-performer (Node 20, zero deps, single-concurrency queue, TASK.md + attachment download + optional git clone, claude -p --output-format json --dangerously-skip-permissions, simulated result when CLI unavailable, artifact serving, HMAC-signed callbacks with retry); compose profile mocks w/ healthchecks + ${HOME}/.claude mounts; contract tests pass live (WP ran real Claude Code via mounted host creds). UFW rule added for container→host callbacks.