Add local setup using uv package manager

[jquentino]

Motivation

The Google Colab Free plan has a very limited GPU quota (#201) that makes it difficult to complete the labs effectively. Once the quota is exhausted, users are forced to wait extended periods before regaining GPU access. This PR adds support for running labs locally using the uv package manager, allowing students to:

• Run labs locally with their own GPU hardware without quota restrictions
• Work continuously without GPU access interruptions
• Use familiar local development tools (VS Code, Jupyter)

Changes

• Added pyproject.toml

  • All required dependencies and versions
  • PyTorch CUDA index configuration
  • Python version constraints

• Updated README with new "Running the labs locally with uv" section that explains:

  • How to install dependencies using uv sync
  • How to switch CUDA versions by editing the PyTorch index URL
  • How to launch Jupyter from the uv environment

Notes for Reviewers

• Only added dependencies for PyTorch labs.
• The default CUDA version is set to 12.6, but can be easily changed in ´pyproject.toml´
• Dependencies are pinned to minimum versions to maintain compatibility
• Python version is set to 3.12 to match the current Colab environment