test: expand E2E smoke tests with 14 new CLI verification tests
Add comprehensive black-box smoke tests covering command routing,
graceful error handling, headless mode validation, and help completeness.
New tests:
- Command routing: headless --help, sessions --help
- Flag aliases: -v (--version), -h (--help)
- Error handling: no-TTY clean exit, unknown flags resilience
- Headless mode: missing .gsd/ dir, missing --context, invalid/negative --timeout
- Help completeness: all subcommands listed, all key flags listed
- Edge cases: --version ignores trailing args, headless positional help
All tests run without API keys and use temp directory isolation.
2d037249c4