feat(cli): add /thinkback to review model thinking blocks

[emal-avala]

Summary

Adds `/thinkback` — surfaces the model's thinking blocks from a recent turn so users can review reasoning after the fact.

• `/thinkback` — most recent turn's thinking blocks
• `/thinkback ` — nth most recent turn (1 = latest)
• Walks `state.messages` for `ContentBlock::Thinking`; skips user messages and non-thinking blocks
• Errors cleanly when there's no thinking in the session or `n` exceeds the count

Why

Extended thinking is enabled in agent-code (interleaved-thinking beta flag), and blocks are persisted in conversation history. But once they scroll off the terminal, they're effectively invisible — there was no way to review why the model went down a specific path. This command closes that loop.

Test plan

• `cargo fmt --all` clean
• `cargo check` passes
• `cargo clippy --no-deps` clean
• `cargo test --bin agent commands::tests::thinking_walker_skips_user_messages_and_non_thinking_blocks` passes (verifies it picks thinking blocks from assistant messages only and skips messages without thinking)
• Manual: after a turn with thinking, `/thinkback` prints the thinking text
• Manual: `/thinkback 99` in a short session prints a clear "only N turn(s)" message
• Manual: `/thinkback` in a fresh session prints "No thinking blocks in this session yet."

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.