[seidewitz]

This PR revises the update to default multiplicity implementation made in PR #815.

The main purpose of the changes in PR #815 were to move the implementation of default multiplicity from the transformation adapters to the post-processing adapters. However, it also removed one of the criteria for adding a default multiplicity to a usage: "The usage does not have any explicit owned subsettings or owned redefinitions that have no owning type" (as it was previously implemented).

Subsequently, testing on the library and example models included in a release indicated that this criteria change had a significant impact, resulting in several cases in which explicit multiplicities needed to be added to redefining features (specifically when the original multiplicity lower bound was greater than 1). This is a negative impact for the modeler, because it requires looking up the redefined feature and copying its multiplicity.

So, this PR adds back a limited version of the criterion removed in PR #815: "The usage does not have any explicit owned redefinition relationships." Leaving non-redefinition subsetting out of this criterion (i.e., giving default multiplicities to subsetting but non-redefining features) avoids situations such as the following:

attribute m1 : ISQ::MassValue; // Multiplicity 1..1
attribute m2 :> ISQ::mass;     // Multiplicity 0..* (if no default multiplicity)