Auto-mode range and settings feature idea

### Describe the feature or problem you'd like to solve

I'd like Auto-mode to have more local wiring for users. For example, I like the concept of auto-mode but I would like to set a minimum strength model a maximum strength and bias. Like bias towards stronger models.

### Proposed solution

Auto mode should have a config entry which has some basic levers. I imagine at least these 3:

Lowest selectable model.
Cost saver - intelligence bias slider value (0 meaning lean towards the cheapest model for a given task - 1 being choose the most almost always)
A short list of models that I am willing to use. For example, why is it still usign Haiku when it hasn't been updated since 2025 and we have many small models that are much more capable.

### Example prompts or workflows

_No response_

### Additional context

_No response_

The issue does not identify files, tests, or entry points. Start by locating Auto-mode configuration and model-selection behavior in the CLI, then clarify the supported settings and model list before defining tests that establish the expected selection behavior.

Written by the indexing model from the issue text.