Draw digits interactively in JLab/SMStudio for migration challenge

[athewsey]

Issue #, if available: #3

Description of changes:

Enables interactively drawing digits and classifying them with trained model for the MNIST challenge - which we used to support only for plain Jupyter environments (where, unlike JupyterLab/SMStudio, plain JS+HTML widgets could be embedded and directly call functions on the Python kernel).

Status:

• [X] Tested on SageMaker Studio JLv3
• [X] Tested on SageMaker Notebook Instance JLv3 (platform notebook-al2-v2)
• [X] Tested on JupyterLab v1
  • This seems to work but the extension install process is a bit different: While pip install ipycanvas from the JupyterServer's host environment is sufficient for JLv3, jupyter labextension install ipycanvas seems to deliver usable results for JLv1 and v3. Both cases seem to require restarting the JupyterServer to behave properly, and v1 a jupyter lab rebuild may also be needed.
  • Overall, while the code itself seems to work on JLv1 when ipywidgets and ipycanvas extensions are installed, setup might not be quite as clean.
• [ ] Ported to PyTorch alternative

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[athewsey]

Merging this without PyTorch port for now, as it's likely we'll work on bringing the PyTorch & TensorFlow alternatives into same exercise folders soon which would involve re-aligning them anyway.