CodeRabbit Plan steers coding agents

This is the right place to attack agentic coding failures: not after a messy PR, but before the first line of code. CodeRabbit is moving from reviewer to planning copilot, which is where the leverage is.

• –Plan turns messy tickets into reviewable, machine-readable prompts instead of one-off editor chats
• –Repo and issue context should cut down on guesswork and keep plans tied to real codebase constraints
• –Handoff to external agents makes CodeRabbit feel more like a planning layer than a pure review bot
• –The upside is less rework and fewer hallucinated assumptions; the risk is that weak tickets still produce weak plans