Add configurable headless defaults

[RobertTLange]

What changed

• Added ~/.headless/config.toml support for per-agent default model and reasoning_effort values.
• Added a packaged config.toml.example and README docs for setup, supported keys, and precedence.
• Updated the changelog with all commits since v0.2.0, including usage accounting, provider defaults, Gemini trust handling, OpenCode output extraction, Pi provider/model specs, Cursor model/effort variants, and config defaults.

Why

• Lets users set stable defaults once instead of repeating model and reasoning flags for every headless run.
• Keeps CLI flags and existing Codex/Pi environment overrides as higher-priority escape hatches.

Verification

• npm run check
• npm pack --dry-run
• npm install -g .
• headless opencode --prompt "hello" --print-command
• Temp HOME smoke with [agents.opencode] model = "openai/gpt-5.5" and reasoning_effort = "high"