hermes - 💡(How to fix) Fix feat: add pre_generate hook to VideoGenProvider for quota/cost checks before dispatch

Problem

There is no hook point before generate() is called by Hermes. Implementing pre-generation checks (quota, rate-limits, cost gates) requires embedding the check logic inside generate() itself, which adds latency and couples quota logic to generation logic.

Proposed Change

Add an optional pre_generate() hook to VideoGenProvider:

def pre_generate(self, prompt: str, model: str, **kwargs) -> Optional[dict]:
    """Called before generate(). Return a dict to short-circuit (use as generate() result). Return None to proceed normally."""
    return None

Motivation

Mother's Luma plugin needs to check the brand's monthly budget before dispatching to Luma. Currently implemented as an MCP call inside generate(). A native hook would make this cleaner and lower-latency, and remove the dependency on Mother's MCP server being available at generation time.