[Bug Fix] Make execution subagent model exp based instead of simple.

[vikramnitin9]

The execution subagent model is not set as experiment-based, which means that our flights are not using the correct models. Fixed that, and also changed the default execution subagent tool calling limit to 10.

Supercedes #308570.

@bhavyaus - tagging you for approval.

[Copilot]

Pull request overview

Fixes execution subagent configuration so model selection is experiment-based (ensuring flights pick up the intended model), and updates the default execution subagent tool-calling limit.

Changes:

• Switch chat.executionSubagent.model to an experiment-based config and read it via getExperimentBasedConfig(...).
• Increase default chat.executionSubagent.toolCallLimit from 5 to 10.
• Mark execution subagent settings as experiment-driven in the extension configuration (onExp tag).

Show a summary per file

File	Description
extensions/copilot/src/platform/configuration/common/configurationService.ts	Makes ExecutionSubagentModel experiment-based and raises default tool-call limit to 10.
extensions/copilot/src/extension/prompt/node/executionSubagentToolCallingLoop.ts	Reads execution subagent model via experiment-based configuration lookup.
extensions/copilot/package.json	Updates defaults/tags for execution subagent settings to reflect experiment-based behavior and new limit.

Copilot's findings

• Files reviewed: 3/3 changed files
• Comments generated: 0