[App] Allow custom presets in the model picker power slider

What version of the Codex App are you using?

Codex App 26.707.51957

What platform is your computer?

macOS arm64

What feature would you like?

Allow users to define the presets shown in the Codex App's Advanced model picker / power slider.

The app already persists related desktop settings in ~/.codex/config.toml, for example:

[desktop]
enabled-reasoning-efforts = ["low", "medium", "high", "xhigh", "ultra", "max"]
show-ultra-in-model-picker-slider = true

It also supports global defaults such as:

model = "gpt-5.6-sol"
model_reasoning_effort = "high"

However, these settings do not let a user choose which model + reasoning combinations appear as slider stops. The current slider uses a fixed curated sequence.

I would like to configure my own ordered list, for example:

[desktop]
model-picker-presets = [
  { model = "gpt-5.6-terra", reasoning_effort = "high", label = "Terra High" },
  { model = "gpt-5.6-sol", reasoning_effort = "high", label = "Sol High" },
  { model = "gpt-5.6-sol", reasoning_effort = "xhigh", label = "Sol Extra High" },
  { model = "gpt-5.6-sol", reasoning_effort = "max", label = "Sol Max" },
  { model = "gpt-5.6-sol", reasoning_effort = "ultra", label = "Sol Ultra" },
]

The exact schema is only illustrative.

Why is this useful?

Different users have different recurring workflows and cost/latency preferences. A fixed slider sequence forces users to open the full picker repeatedly or accept presets they do not use.

Custom presets would make the shortcut genuinely useful while preserving the existing defaults for users who do not configure anything.

Expected behavior

• Keep the existing curated presets as the default.
• Allow users to override the ordered slider stops in config.toml or Settings.
• Each preset should support at least model and reasoning_effort.
• Optional labels would improve clarity.
• Invalid or unavailable presets should be ignored with a visible warning rather than breaking the picker.
• The selected preset should stay synchronized with the full model/reasoning controls.

Related issue

#31968 discusses the current slider's fixed sequence and its treatment of Max and Ultra, but it does not request user-defined presets.