Add separate effortLevel configuration for advisor model

Summary

Currently, effortLevel in settings.json applies only to the executor model.
There is no way to configure the effort level for the advisor model separately.

Current Behavior

{
  "model": "claude-sonnet-4-6",
  "advisorModel": "claude-opus-4-7",
  "effortLevel": "high"
}

The advisor model's effort level is managed server-side dynamically.
Users cannot tune it independently.

Desired Behavior

{
  "model": "claude-sonnet-4-6",
  "advisorModel": "claude-opus-4-7",
  "effortLevel": "high",
  "advisorEffortLevel": "xhigh"
}

Use Case

• Executor (Sonnet): high — fast iteration, lower cost
• Advisor (Opus): xhigh — maximum reasoning for critical decisions

Opus 4.7 defaults to xhigh, but there's currently no way to confirm or override this per-project. Separating the two settings would give teams explicit, auditable control over cost vs. quality tradeoffs.

Related

• Not a duplicate of #35108 (that request is about per-tool effort, not advisor/executor split)