v0.7.2 — Terminal Display, Execution Timeout & Streaming Token Usage

What's New

Terminal Display (Claude Code-style UI)

• Rich-based terminal rendering with display=True (default)
• ● prefix for code execution, ⎿ for output, live Markdown streaming
• Token usage summary on completion (done 6.5s · 983 tokens)
• Works with both run() and stream_events()

Execution Timeout

• max_exec_timeout parameter to limit code execution time
• Thread-based timeout handles sync-blocking code (network, DB, time.sleep)
• IPyKernelRuntime sends SIGINT for real interruption
• LLM receives timeout feedback and adjusts strategy
• System prompt auto-guides LLM to use timeout params in network/DB calls

Streaming Token Usage

• StreamResponse abstraction tracks token usage during streaming
• stream_options: include_usage for OpenAI-compatible endpoints
• Token counts displayed in terminal summary

Other Changes

• Removed Logger, replaced by display system
• system-reminder injects current datetime per request
• dill.dumps(recurse=True) fixes serialization with Rich globals
• _extract_token_usage deduplicated to Model base class
• requires-python bumped to >=3.12 (IPython 9 requirement)
• 9 new timeout unit tests (mock-based, no LLM needed)