Add OpenCode CLI Provider #534

As an OpenCode user, I want SkillSpector to support OpenCode as an LLM provider so I can run semantic security scans on my skills without leaving my OpenCode session.

The gap

OpenCode users currently get static-only scans because SkillSpector lacks an opencode_cli provider: there is no way to set SKILLSPECTOR_PROVIDER=opencode_cli, so llm_available stays false and the semantic analyzers are skipped.

What I am asking for

Please consider OpenCode a supported CLI provider, on par with the existing Claude/Codex/Gemini CLI providers: selectable via SKILLSPECTOR_PROVIDER=opencode_cli, using the local opencode binary's own login (no API key handling), with the prompt delivered via stdin and the reply parsed from opencode run --format json events. Done when an OpenCode user can run a semantic scan with the CLI-default model (or a SKILLSPECTOR_MODEL override) and get llm_available: true with findings carried by the semantic analyzers.

This could be achieved by

• Adding providers/opencode_cli/provider.py and __init__.py mirroring the codex_cli pattern
• Registering the provider in _agent_cli.py with argv builder (opencode run --format json --pure), JSON-event parser, and opencode auth list auth check
• Wiring selection via SKILLSPECTOR_PROVIDER=opencode_cli in providers/__init__.py and inference_usage.py
• Updating docs trio: README provider table, .env.example, docs/DEVELOPMENT.md
• Registry scope kept minimal: free-model token-budget catalog split out of this change (separate chore), so the provider diff stays wiring-only
• TDD test suite in tests/provider/test_opencode_cli.py covering argv, auth, parser, wiring, registry label
• Live verification: nonce round-trip proves stdin piping works despite no documented flag