feat(sdk): add Python SDK with full feature parity

[saurabhjain1592]

Summary

Introduces the AxonFlow Python SDK with full feature parity to the Go SDK.

Key Features

• Async-first design with sync wrappers via AxonFlow.sync()
• Full type hints with Pydantic v2 models
• Gateway Mode for lowest-latency LLM calls
  • get_policy_approved_context() for pre-checks
  • audit_llm_call() for compliance logging
• LLM Interceptors for transparent governance
  • wrap_openai_client() - OpenAI integration
  • wrap_anthropic_client() - Anthropic integration
• MCP connector operations (list, install, query)
• Multi-agent planning (generate, execute, status)

Architecture

sdk/python/
├── axonflow/
│   ├── __init__.py          # Public API exports
│   ├── client.py            # Main AxonFlow client (813 lines)
│   ├── types.py             # Pydantic v2 models (213 lines)
│   ├── exceptions.py        # Exception hierarchy (103 lines)
│   ├── interceptors/        # LLM provider interceptors
│   │   ├── openai.py        # OpenAI wrapper
│   │   └── anthropic.py     # Anthropic wrapper
│   └── utils/               # Utilities (cache, retry, logging)
├── tests/                   # Comprehensive test suite
├── examples/                # Usage examples
├── pyproject.toml           # Package configuration
└── README.md                # Documentation

Quality Standards

• Linting: ruff check passes
• Type checking: mypy strict mode compatible
• Test coverage: 95%+ target
• Python support: 3.9, 3.10, 3.11, 3.12
• Dependencies: httpx, pydantic v2, tenacity, structlog, cachetools

Usage Examples

Async (recommended):

async with AxonFlow(agent_url="...", client_id="...", client_secret="...") as client:
    response = await client.execute_query("user-token", "What is AI?", "chat")

Sync:

with AxonFlow.sync(agent_url="...", client_id="...", client_secret="...") as client:
    response = client.execute_query("user-token", "What is AI?", "chat")

Gateway Mode:

ctx = await client.get_policy_approved_context(user_token, query, data_sources=["postgres"])
if ctx.approved:
    llm_response = await openai.chat.completions.create(...)
    await client.audit_llm_call(ctx.context_id, summary, "openai", "gpt-4", usage, latency)

Test plan

• CI checks pass (lint, type check, tests)
• Python 3.9, 3.10, 3.11, 3.12 compatibility
• Test coverage >= 90%
• Package builds successfully
• Examples run without error

[saurabhjain1592]

Closing: PR should be created in axonflow-enterprise first, then sync to OSS.