Simplify NI-FGEN example and remove its use of numpy and scipy

[ni-jfitzger]

• This contribution adheres to CONTRIBUTING.md.

• I've updated CHANGELOG.md if applicable.

• [ ] I've added tests applicable for this pull request

What does this Pull Request accomplish?

• Eliminates our use of numpy and scipy from the NI-FGEN example, and slightly simplifies the waveform creation, to keep the focus on the usage of NI-FGEN.
  • The new waveforms will only be a single cycle. Previous waveforms were an arbitrary 1.x cycles, because the phase ended at 10, which is greater than 2PI and not even a multiple of 2PI.
• The Triangle waveform in the example was actually a sawtooth. We will now refer to it as such.
  • I could have added an actual triangle wave, but doing so with 0 offset would have been complicated enough that I worried I felt it would distract from the focus on the API usage. 3 or 4 lines, probably, but still the most complicated of the waveforms.
• Scipy has stopped releasing Windows 32-wheels for scipy 1.8.0+ and for Python 3.10 and later. This is breaking nimi-python testing for PRs, when we try to add Python 3.10 support. This change fixes that.
  • Side note: building the scipy Windows 32 wheel would be non-trivial.

List issues fixed by this Pull Request below, if any.

• Fix nimi-python uses scipy which keeps us from adding support to Python 3.10 #1811

What testing has been done?

• Hand-tested updated functions, because they are part of the example code and not unit-tested.
  • Examined the output (in list form) from the updated functions and they all look correct to me.
  • I plotted the noise waveform and it looked good.
  • I generated a 1000 point noise waveform with the same algorithm and I think the max absolute value was around 0.8. It never exceed the bounds of [-1, 1].