Add runtime requirements

[lwneal]

Simply copies kornia, streamlit-drawable-canvas, and invisible-watermark from the conda-style environment.yaml to the pip-style requirements.txt.

Tested with the following:

    git clone https://github.com/Stability-AI/stablediffusion/
    cd stablediffusion
    python -m venv venv
    source venv/bin/activate
    pip install -U pip
    pip install -r requirements.txt
    python scripts/txt2img.py --prompt "our galaxy itself contains a hundred billion stars" --ckpt ../768-v-ema.ckpt --config configs/stable-diffusion/v2-inference-v.yaml --H 768 --W 768  --n_samples 1

on Ubuntu 22 with Python 3.8, or similar, on a machine with CUDA already installed.

[woctezuma]

This would fix #31. 👍