feat: add support for Gemini CLI

Description

Gemini CLI (https://github.com/google-gemini/gemini-cli). Tier-2 target under epic #66 alongside Codex.

Gemini CLI is a "universal" agent: it reads .agents/skills/ directly, so once the repo is published to the open standard (skills.sh), the workflow skills land in Gemini CLI for free. Remaining work is the hook and MCP, plus the subagent-degradation decision.

Scope

• Skills — verified to load from .agents/skills/.
• Hook — port recommend-plugins to Gemini CLI hooks. Gemini's hook system mirrors the Claude Code JSON-over-stdin contract and exit-code semantics (same PreToolUse / PostToolUse / UserPromptSubmit / Stop events) and ships migration tooling that converts existing Claude Code hooks.
• MCP — wire context7 via Gemini CLI settings.json.
• Subagents — Gemini CLI has full subagent support (since v0.36): custom agents are Markdown + YAML frontmatter in .gemini/agents, with isolated context, curated tools, MCP, and parallel dispatch. Format is nearly identical to our existing agents/*.md, so the review/plan/research panels port with minimal change.
• CI — loader job asserting skills load in Gemini CLI.

Docs: https://geminicli.com/docs/hooks/reference/

Requirements

• Skills load in Gemini CLI via .agents/skills/
• recommend-plugins hook ported via Gemini's Claude-hook migration tooling
• context7 MCP wired via settings.json
• Agent definitions placed in .gemini/agents (md+frontmatter); review/plan panels run natively
• Gemini CLI loader CI job added

Additional Context

Part of epic #66. Tier 2 / wave 1 alongside Codex (#226) and OpenCode (#229).