Subagent ignores reasoningEffort setting from agent config

[johnnyapu15]

Using GPT 5.2, primary agents (build, plan) respect reasoningEffort config, but subagents (explore, general) ignore it.

Root cause In task.ts:145-160, when calling SessionPrompt.prompt() for subagents, agent.options (which contains reasoningEffort) is not passed.

Primary agents work because llm.ts:106 merges input.agent.options into LLM options.

Fix Add options: agent.options to the SessionPrompt.prompt() call in task.ts.