Blocks that cannot run as printed: fill-in-the-blank exercises, a singular-matrix demonstration, and a plotly scatter without x and y

[kgdunn]

Executing every code-block:: python in the book (#274) turned up a handful that fail for reasons unrelated to process-improve:

• univariate-review/normal-distribution-and-checking-for-normality.rst:296-305 and :324-336, and least-squares-modelling/covariance-and-correlation.rst:158-180: fill-in-the-blank exercises (____). These are meant not to run; they need a .. code-check: skip marker with the reason so the checker leaves them alone and a reader knows they are exercises.
• design-analysis-experiments/fractional-factorial-designs/half-fractions.rst:69-84: inverts a singular X'X on purpose ("Calculate the inverse (it will fail!)"). Catch numpy.linalg.LinAlgError and print the message so the demonstration runs and shows what it means to show.
• process-monitoring/process-monitoring-exercises.rst:812-830: .plot.scatter() is called without x= and y= after setting the plotly backend, which requires both.
• product-development-product-improvement/soft-sensors.rst:197: a >>> doctest-style block; the rest of the book prints and echoes the result in a comment.