WIP: Improve Task Runner Examples and Documentation

[psfoley]

This PR:

• Makes changes to the PyTorchTaskRunner to wrap user train_ and validate_ functions. The impact of this is that for basic examples (including the quickstart), users don't need to understand or configure tensorkeys
• Adds on to the @kta-intel and @nammbash's quickstart and README content to show how to:
  1. Run a basic experiment for federated training
  2. Customize the deep learning model
  3. Customize training and validation tasks
  4. Configure the dataloader
  5. And rerun another experiment with those changes
• Several other small changes to the documentation

Items remaining before removing WIP tag:

• [x] Verifying all other PyTorch TaskRunner API examples (i.e. torch_cnn_histology, etc.) are compatible with base changes
• [ ] Additional documentation updates