[Feat]: Add Go GenAI SDK reference implementation and CLI for Agentic Video

### Is your feature request related to a problem? Please describe.

Currently, the `gemini/agentic-video` samples are only available as a Python Jupyter notebook (`intro_agentic_video.ipynb`). Developers building production services, microservices, and command-line tools in Go need idiomatic reference implementations, clean client factories, and patterns using the official Google GenAI SDK (`google.golang.org/genai`).

### Describe the solution you'd like

Contribute a Go reference implementation and CLI application in `gemini/agentic-video/gemini-agentic-video-go`:
- **Core Agentic Video Patterns:** Single-video dynamic timeline navigation, multi-video comparative synthesis, and stateful multi-turn dialogues maintaining video context.
- **Go SDK Reference & Engineering Guide:** Decoupled packages (`internal/runner`, `internal/client`, `internal/telemetry`) that can be embedded into headless backend services.
- **Concurrent Benchmarking:** Side-by-side mode shootout (Agentic vs. Static 1-FPS) and multi-model comparison (`gemini-3.8-flash`, `gemini-3.7-flash`, `gemini-3.6-flash`) using Go's native goroutines and live telemetry.
- **Comprehensive Documentation:** Complete developer and architectural guides covering token accounting shifts (dynamic frames accounted in thoughts tokens), content part ordering rules, and thinking configurations.

### Describe alternatives you've considered

Using only Python notebooks or raw REST API curl calls, which lacks idiomatic Go type safety, concurrency patterns, and production engineering guidance for the Go developer community.

### Additional context

A PR with the complete implementation, unit test suite (race detector clean), and documentation has been prepared on a feature branch.