Task: Add # may vary tag to random output, and automate line length issues

[bmwoodruff]

Description:

The current reviewtools.py file process_text_block does not check for

• lines longer than 79
• random output

This causes two problems:

1. When running spin lint, the generated examples will fail if longer than 79 characters (thanks @jud-sdev for help in Review: AI Gen Examples for ma.dot, ma.cumsum and ma.prod #87).
2. When running python tools/refguide_check --doctests, random output will fail (thanks @ogidig5 for help in Review: AI Gen Examples for numpy.linalg.svdvals function #86).

I'd like to automate much of this.

• I have to find a good way to reformat python code (I'm sure there is a package out there that will reformat code to stay within a specified length. I already know how to use a textwrapping function in the commit_message_creator function, which will handle regular text just fine. The issue will be formatting python code (and I'm sure it's been done).
• I'll play with a few options for detecting random (search for rand, then expand it to np.rand or other things till it captures exactly what's needed.)
• Update reviewtools.py with two new functions to do this.

Acceptance Criteria:

• Random output is appropriately tagged in reviewtools.py
• Lines longer than 79 characters are appropriately split into shorter lines, resulting in still valid python code.

[bmwoodruff]

Looks like black is designed to format long code, so that may work. I'll have to substract the indent (and may be the >>>) from the length, but I think this will work for getting input code lengths shrunk to under 79 characters. I know they're planning to increase the limit to 88 soon. If they get that done before we're done, then we swap a number.

The package textwrap will format long strings. This works for the headers between code blocks. If they are multiline, then combine them into a string before wrapping them and then writing them.

I'm not yet sure about formatting the output cells to be 79 or less.