feat(providers): Add OpenAI provider adapter

[sunilp]

Summary

Add an OpenAI-compatible provider adapter so users can use GPT models with Jam.

Details

Implement ProviderAdapter from src/providers/base.ts for OpenAI's chat completions API.

Requirements

• Create src/providers/openai.ts implementing the ProviderAdapter interface
• Support streaming via SSE (stream: true)
• Register in src/providers/factory.ts
• Add unit tests in src/providers/openai.test.ts
• API key should be read from keychain or JAM_API_KEY env var

Reference

• OpenAI Chat Completions API
• Existing Ollama adapter: src/providers/ollama.ts

Getting Started

See Adding a New Provider in the README.

[sunilp]

✅ Implemented and verified — closing

Merged in #11 (commit 1c908d9).

Delivered:

• src/providers/openai.ts — full ProviderAdapter: SSE streaming, chatWithTools, listModels, validateCredentials
• API key from OPENAI_API_KEY env var or profile apiKey
• Registered in factory.ts with PROVIDER_AUTH_FAILED / PROVIDER_MODEL_NOT_FOUND / PROVIDER_RATE_LIMITED / PROVIDER_STREAM_ERROR error codes

Test results — 10/10 tests pass (npm test, 160 total):

✓ src/providers/openai.test.ts (10)
  parses SSE stream and yields deltas
  throws PROVIDER_AUTH_FAILED on 401
  throws PROVIDER_MODEL_NOT_FOUND on 404
  throws PROVIDER_RATE_LIMITED on 429
  throws PROVIDER_STREAM_ERROR on non-OK response
  includes system prompt when provided
  validateCredentials resolves when API key is valid
  validateCredentials throws PROVIDER_AUTH_FAILED when key is invalid
  listModels returns sorted model ids
  throws PROVIDER_AUTH_FAILED when no API key available