Your personal operating system for career transition.

Implement a CostAwareRouter that selects the cheapest model capable of handling a request:

1. Takes a list of ModelOption objects: name: str, cost_per_1k_tokens: float, max_context_tokens: int, supports_json_mode: bool
2. route(prompt_tokens: int, requires_json: bool = False) -> ModelOption — returns the cheapest model that:
   • Has max_context_tokens >= prompt_tokens + 512 (leave headroom for completion)
   • If requires_json=True, only consider models where supports_json_mode=True
3. Raises ValueError if no model can handle the request
4. estimate_cost(model: ModelOption, input_tokens: int, output_tokens: int) -> float — static method returning estimated cost in USD
5. Track routing decisions and expose routing_stats() -> dict with counts per model name

Test with three models: a cheap small model, a mid-tier model, and an expensive large model. Verify that simple requests route to the cheap model and long-context requests route to the only model that can handle them.