Some TST panel feedback

[tylerjereddy]

Just a few comments from inspecting the codebase and poking around a bit in case they are helpful--some of them may already be resolved and just need to be clarified somewhere perhaps.

• the testing framework seems to be split between conventional
  pytest and a strange-looking test/unit/run.py test harness--are
  both really needed? why?
• test infrastructure using assert math.isclose/np.isclose instead
  of proper NumPy test assertions that provide more detailed
  failure feedback
• when I run python -m pytest locally in py3.13 there is one
  failure (test_multi_table_distribution_sample -- E TypeError: sample_multi_table() got an unexpected keyword argument 'scale')
• test line coverage is reported as ~33 % when I check locally:
  python -m pytest --cov=. --cov-report=html; is that correct?
• formal performance benchmarking with asv might make sense?
• motivate the need to use numba--often working with non-rectangular
  data structures? (vs. using the array API standard?)
• consider making immmutable releases on GitHub (DOE cares
  about supply chain security, and it is just a checkbox really)