refactor(ask-a-sailor): swap OpenAI for any-llm provider abstraction

[Copilot]

AskASailorAgent hard-codes OpenAI(), locking the agent to a single vendor. This swaps in mozilla-ai/any-llm so clubs can switch LLM providers via env var.

Agent refactor

• Replace from openai import OpenAI → from any_llm import completion, embedding
• Remove self.client = OpenAI(); call module-level completion() / embedding() directly
• Model resolved from LLM_MODEL env var, falling back to gpt-4o-mini

# Switch providers without code changes
LLM_MODEL=gpt-4o-mini              # OpenAI (default)
LLM_MODEL=mistral-small-latest     # Mistral
LLM_MODEL=anthropic:claude-3-haiku # Anthropic
LLM_MODEL=ollama:llama3            # Local/offline

Dependencies

• openai>=1.20.0 → any-llm-sdk[openai]>=1.10.0 in both root and package requirements (openai remains as transitive dep)

Config & CI

• New .env.example documenting LLM_MODEL and provider API keys
• LLM_MODEL added to CI test env

Tests

• test_agent_answer_with_mocked_any_llm — end-to-end with mocked completion/embedding
• test_agent_respects_llm_model_env — verifies env var propagates through to the completion call

All 30 tests pass (28 original + 2 new), 6 integration tests skipped as before.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>[ask-a-sailor] mozilla-ai/any-llm provider abstraction: swap LLMs without code changes</issue_title>
<issue_description>Currently AskASailorAgent hard-codes OpenAI. mozilla-ai/any-llm (github.com/mozilla-ai/any-llm) provides a single interface across OpenAI, Anthropic Claude, Mistral, and local models (Ollama) with one import swap. This is critical for: (1) cost flexibility -- switch to Mistral 7B for lower-cost club deployments, (2) offline/local mode -- llamafile for clubs with no cloud budget or data privacy requirements, (3) funder credibility -- not locked to any single vendor. Work: (1) pip install any-llm, refactor AskASailorAgent to use any_llm.get_llm(model=os.environ['LLM_MODEL']) instead of OpenAI() directly, (2) add LLM_MODEL env var to .env.example and CI, (3) verify tests still pass with mocked any-llm client, (4) document the swap in README.</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes [ask-a-sailor] mozilla-ai/any-llm provider abstraction: swap LLMs without code changes #21

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.