/tangent feedback

I've been playing around with the experimental `/tangent` feature and wanted to provide some feedback. First off, it's quickly become part of my normal flow, so conversation branching is definitely super useful, so thanks for that! However, one issue I have with it is that I need to explicitly choose when to start a tangent session and I don't always know if I'm going to need one or not. I'm sort of using it as a way to hedge in case my request results in more of the context window being taken up than I want. For example, I'll start a tangent if I'm going to have Q do research the codebase just in case it ends up having to search through many many files and I want a way to reset back to a previous checkpoint. However, sometimes it doesn't end up using much of the context window, and now if I run `/tangent` again, it'll forget what it just did.

I think the better experience would be, rather than me having to explicitly run `/tangent` to create a checkpoint, if I could just walk back in the conversation history and choose to reset back to that point. That basically gives me the same value, but I get to make the decision when I have enough information to know if I actually want to reset to a previous checkpoint or not.