ENH: first beta release optimizations and fixes

Before releasing the first beta, we should do the following:

To-do for v1.0.0b1

Optimizations

• Optimize the nearest_preferred_unit algorithm to run in ~3 ms or less and use less memory. We will still do it in pure Python and see if it needs to be written in C to get more speed.
• Create an artificial benchmark that simulates N widgets on screen and simulates normal interactions and operations, like updating the values of the widgets and performing unit conversions. This benchmark should measure the time from operation start to operation end for each widget and the memory used by the entire process. This will help to troubleshoot performance issues with PyMR.

Fixes

• Solve the problem with pytest on the CircleCI pipeline: when running pytest on the pipeline environment, it raises a segmentation fault and fails. This doesn't occur on local environment or in tox with xvfb-run.

Features

• Nothing yet.

Documentation

• Nothing yet.