Add new 'opencode provider' command suite for managing custom model providers:

- provider list: Display all custom providers from opencode.json with details
  including provider name, base URL, and available models
- provider add: Interactive wizard to add new custom providers with
  prompts for provider ID, name, base URL, SDK package, and models
- provider remove: Interactive removal of custom providers with
  confirmation prompt

Features:
- Easy management of custom OpenAI-compatible providers (Ollama, LM Studio, etc.)
- No need to manually edit opencode.json files
- Full integration with existing 'opencode models' and 'opencode run' commands
- Supports multiple models per provider
- Validates provider IDs and URLs during input

Also added opencode.json to .gitignore as it's a local configuration file.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

- Add -mcdev suffix to all build versions for custom build identification
- Extend version timestamp to include seconds for better build tracking
- Add connection testing and model discovery to provider management
- Enhance TUI sidebar with detailed token breakdown and performance metrics
- Display context usage as "used / total" with percentage
- Show latency and tokens/sec performance stats

These changes make it easier to identify custom builds and provide
better visibility into model usage and performance during sessions.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>