Update gpindices shifts multiple times when you add in a new parameterized SPAM layer.

[nkoskelo]

No description provided.

[rileyjmurray]

Thanks for this fix, and for writing the tests!

I have two requests.

1. Please make the tests instance methods of a new class, TestRebuildParamvec, that subclasses unittest.TestCase.
2. Please update the PR opening comment to explain the purpose of the PR. You, me, and the other pyGSTi devs will know the purpose for the near future, but we'll inevitably forget.

I'll note that a search through the pyGSTi codebase shows that ModelMember.shift_gpindices is called only in two places: once recursively in shift_gpindices itself, and once in OpModel._rebuild_paramvec. The latter wasn't passing in any value for the "memo" argument in ModelMember.shift_gpindices.

[coreyostrove]

We'll also need to track down what is going on with the failing unit tests. These look substantive, so there may be some unexpected consequences being caught. Do these fail locally on your system?