Make physical units check trigger for `_GeneralVarData.set_value` more consistent

[shermanjasonaf]

Summary

As of #3077 (and, in particular, cfa6ff49f), invoking _GeneralVarData.set_value() on an variable object with units to a dimensionless numpy type (such as numpy.float64) may result in a UnitsError, but this appears to depend on on whether or not numpy types were added to native_numeric_types before or after set_value() was invoked.

Steps to reproduce the issue

Before cfa6ff49f

>>> import numpy as np
>>> import pyomo.environ as pyo
>>> m = pyo.ConcreteModel()
>>> m.v = pyo.Var(units=pyo.units.m)
>>> m.v.set_value(np.float64(1))
>>> # (no error)

After cfa6ff49f

>>> import numpy as np
>>> import pyomo.environ as pyo
>>> m = pyo.ConcreteModel()
>>> m.v = pyo.Var(units=pyo.units.m)
>>> m.v.set_value(np.float64(1))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "...pyomo/pyomo/core/base/var.py", line 397, in set_value
    val = units.convert_value(
  File "...pyomo/pyomo/core/base/units_container.py", line 1384,
 in convert_value
    raise UnitsError(
pyomo.core.base.units_container.UnitsError: Cannot convert dimensionless to m. Units are not compatible.
>>> m.v.set_value(np.float64(1))
>>> # (no error on second attempt; numpy numeric types registered before exception raised)

Error Message

See previous section.

Information on your system

Pyomo version: 6.7.1dev0 (before and after cfa6ff49f)
Python version: 3.9.17
Operating system: Ubuntu 20.04
How Pyomo was installed (PyPI, conda, source): source
Solver (if applicable):

Additional information

Other dependencies: numpy 1.23.1, Pint 0.19.2