Agent configuration needs an effort and length configuration

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

I can only define which model an agent should be using, but cannot configure anything with that model such as effort level and context length.

### Proposed solution

In [custom agents configuration](https://docs.github.com/en/copilot/reference/custom-agents-configuration) there should be a configuration for effort (like claude code has) and context length.

This would allow me to make research agents that can take advantage of higher reasoning and context length vs task agents who don't necessarily need the added effort and window.

### Example prompts or workflows
Sample agent configuration I'd like:
```
---
name: sample-research-agent
description: "Agent to do some deep research in a large codebase."
tools:
- "*"

# Proposed: structured model configuration block
model: claude-opus-4.7 # base model (decoupled from effort/context)
effort: high # low | medium | high | xhigh (claude code supports this already)
context_length: long # default | long
---
```

Also changes to `task` tool call to launch subagents:
```
// Current (only knob is a suffixed model ID)
{
"agent_type": "general-purpose",
"model": "claude-opus-4.7",
"name": "review-agent-d",
"prompt": "..."
}

// Proposed (orthogonal, composable)
{
"agent_type": "general-purpose",
"model": "claude-opus-4.7",
"effort": "xhigh",
"context_length": "long",
"name": "review-agent-d",
"prompt": "..."
}
```

### Additional context

_No response_

Beginne mit der Dokumentation zur Konfiguration von custom agents und der im Issue beschriebenen Aufrufform des task tool. Lege fest, wie effort und context_length unabhängig von der Modellauswahl dargestellt werden sollen, und überprüfe anschließend, dass sowohl die Agentenkonfiguration als auch das Starten von Subagents die vorgeschlagenen Werte unterstützen.

Vom Indexierungsmodell aus dem Issue-Text verfasst.