feat: streaming support for query responses

[0xneobyte]

Problem

The query method waits for the complete response before returning anything to the caller. For long-form answers this introduces noticeable latency — the caller receives nothing until generation is fully complete. Every major AI SDK (Anthropic, OpenAI, Gemini) exposes streaming as a first-class feature.

Proposed Behaviour

Add a query_stream method that yields response chunks as they arrive rather than waiting for the full response.

async for chunk in client.query_stream("What is Python?"):
    print(chunk, end="", flush=True)

• Existing query() is unchanged
• query_stream() returns an async generator
• Citations and metadata returned at end of stream

Files to Modify

File	Change
src/brainus_ai/client.py	Add query_stream async generator method
src/brainus_ai/models.py	Add streaming chunk model
src/brainus_ai/__init__.py	Export new types

Acceptance Criteria

• query_stream() yields text chunks progressively as they arrive
• Existing query() behaviour is unchanged
• Citations accessible at end of stream
• Stream can be cancelled mid-way without errors
• Full type hints on all new methods and models