[UX] Adjusting thinking budget should *not* be done by including word phrases in user messages

There should be an explicit /think command or special hotkey that can be used to enable/disable the thinking budget.

In my opinion it's the wrong UX to control this through phrases in my message. I don't want to find a way to sneak ultrathink in my message to unlock higher thinking budgets - this also confuses the model.

Additionally, there's no phrase to later disable extended thinking after I decide I'm done with the hard problem that required extended thinking.

There's a myriad of additional reasons why this is bad idea, for example:

1. It's not possible to know about the secret thinking phrase functionality without reading about it in the documentation first.
2. It's uncertain which phrases map to what token budget.
3. It doesn't work well cross-languages.

Considering the desired workflow and design of the hybrid reasoning model (i.e. turn thinking on when you want need it, then off when you don't need it), IMO it makes total sense to have this as a separate command which the user can turn on/off.

_NOTE:_ I know about the MAX_BUDGET_TOKENS option - this is not what I want! I want to enable or disable thinking at specific points along my session - not always on or always off.