Progressive evaluation of the exercise notebooks

1. Upon displaying the exercise, all cells before the first answer cells are evaluated
2. Upon validating the exercise, the remaining cells are evaluated

Features:

• Enables randomization using any feature of the kernel
• Enables displaying cell outputs
• Enables myst-style {eval}... substitutions in Markdown cells
• Fast validation
• Backward compatible with existing exercises

New example:

• examples/progressive.md illustrating the above features

Current limitations:

• Substitution in code cells can only be achieved with Jupylates standard random substitutions
• When validating several times, the kernel is not restarted which is fragile in case of non idempotent code or solutions that could be fetched from later cells.
• Only plain/text outputs are supported (no plots, no widgets).

Additional feature:

• Kernel preheating

Improved py_modulo_result.md example:

• Illustrate how to avoid INPUT by using _
• Illustrate that BEGIN/END SOLUTION is not needed when covering the whole cell
• Illustrate how to prevent cheating by using the name of the variable holding the computed value
• Add a warning about remaining way to cheat
• Cleanup metadata; typos