docs: add Google Colab deployment guide for ComfyUI

[Madhavyamjala]

Description

This PR adds a Google Colab deployment badge to the README.md under the Inference section.

Motivation

While Krea-2 is highly capable, the original BF16 weights exceed the 24GB VRAM limit of many consumer GPUs. This Google Colab notebook utilizes the krea2_raw_fp8_scaled model combined with ComfyUI's aggressive CPU-offloading, allowing the community to run the base model entirely on a free 16GB VRAM T4 instance.

Changes Made

• Added an HTML Colab badge to the centralized link block in README.md to match the existing formatting.
• Linked to a maintained notebook pre-configured with the Krea-2 ComfyUI workflow, Qwen3-VL text encoder, and required dependencies.

Testing

• Verified the HTML badge renders correctly alongside the existing <a> tags in the preview.
• Confirmed the Colab notebook executes successfully from a fresh T4 runtime without VRAM OOM errors.