Investigate using a singleton in place of None to trigger _make_default

[braingram]

Currently setting a DataModel attribute to None (which is a common pattern for freeing up memory) results in creation of a default value for the attribute:

stdatamodels/src/stdatamodels/properties.py

Lines 324 to 330 in ee4756f

	def __setattr__(self, attr, val):
	if attr.startswith('_'):
	self.__dict__[attr] = val
	else:
	schema = _get_schema_for_property(self._schema, attr)
	if val is None:
	val = _make_default(attr, schema, self._ctx)

For large arrays (which are likely candidates for removal to free up memory) this results in creation of a large new array with default values.

>>> import stdatamodels.jwst.datamodels as dm
>>> m = dm.ImageModel((1000, 1000))
>>> d0 = m._instance['data']  # avoid the getattr for data
>>> d0.shape
(1000, 1000)
>>> m.data = None
>>> d0 is m._instance['data'] 
False
>>> m._instance['data'].shape
(1000, 1000)

It may be helpful to use a singleton like MAKE_DEFAULT to signify that a default should be made.

>>> import stdatamodels
>>> import stdatamodels.jwst.datamodels as dm
>>> m = dm.ImageModel((1000, 1000))
>>> m.data = stdatamodels.MAKE_DEFAULT  # to trigger a new default for data

This could help to make it more explicit when a default is requested. However it would require changes to code like the following:

stdatamodels/src/stdatamodels/jwst/datamodels/slit.py

Lines 65 to 68 in ee4756f

	self.data = init.data
	self.dq = init.dq
	self.err = init.err
	self.area = init.area