feat: support Codex MCP config discovery

[brandonwise]

Summary

• add Codex TOML parsing for [mcp_servers.<name>] configs alongside existing JSON formats
• auto-discover ~/.codex/config.toml and project .codex/config.toml
• treat Codex auth/env fields as first-class signals and honor enabled = false
• add Codex fixtures, unit coverage, integration coverage, and README support docs

Why now

• Codex MCP is using TOML configs, so agentwise was missing a real config family that users can actually have today: https://developers.openai.com/codex/mcp
• Search results show active Codex setup/debug demand around ~/.codex/config.toml, [mcp_servers.<name>], and MCP reuse/issues: https://developers.openai.com/codex/mcp Codex does not use MCP servers defined in config.toml. openai/codex#3441 https://www.google.com/search?q=Codex+MCP+config.toml+mcp_servers
• AI-agent / MCP security scanning is getting crowded fast, which makes broader config coverage strategically useful now:
  • GitHub trend: https://github.com/snyk/agent-scan
  • Product Hunt: https://www.producthunt.com/products/0xaudit
  • search signal: https://www.google.com/search?q=AI+agent+security+scanner+MCP+Product+Hunt+GitHub
• web_search was unavailable in this runtime, so telemetry fallback used browser Google results, direct fetches, GitHub links, and the TikTok trend pipeline (projects/last-tiktok-trends.md)

Validation

Targeted tests

• cargo test codex -- --nocapture ✅ passed (4 unit + 2 integration tests)
• cargo test remote_server_with_ -- --nocapture ✅ passed (2 auth-rule tests)

Repo-level suite

• cargo fmt --all -- --check && cargo test --all-targets ✅ passed

Lint / build

• cargo clippy --all-targets -- -D warnings ✅ passed
• cargo build ✅ passed
• cargo build --release ✅ passed

Smoke / integration

• cargo run -- scan testdata/codex-config.toml --format json ✅ passed
  • observed configs_scanned: 1
  • observed servers_scanned: 1
  • observed expected AW-005 finding for the insecure sample URL

Docs quality

• node src/cli.js score --file /Users/bwise/.openclaw/workspace/projects/agentwise-pm-20260506/README.md --json (run from projects/humanizer) ✅ passed with {"score":42}