SizingSystem ctor and clone

Issue overview

I wrote some model tests for SizingSystem (there were none), and now seeing some weird things.

Current Behavior

Issue 1

  Model m;
  AirLoopHVAC airLoopHVAC(m);
  SizingSystem sizingSystem = airLoopHVAC.sizingSystem();
  auto sizingSystemClone = sizingSystem.clone(m).cast<SizingSystem>();

  // FIXME: AirLoopHVAC returns the original SizingSystem OR the cloned SizingSystem
  EXPECT_EQ(airLoopHVAC.sizingSystem().handle(), sizingSystem.handle());  // sometimes true, sometimes not
  EXPECT_NE(airLoopHVAC.sizingSystem().handle(), sizingSystemClone.handle());  // sometimes true, sometimes not

Issue 2

  Model m;
  AirLoopHVAC airLoopHVAC(m);
  SizingSystem sizingSystem1 = airLoopHVAC.sizingSystem();
  SizingSystem sizingSystem2(m, airLoopHVAC);

  // FIXME: AirLoopHVAC returns the original SizingSystem OR the new SizingSystem
  EXPECT_EQ(airLoopHVAC.sizingSystem().handle(), sizingSystem1.handle());  // sometimes true, sometimes not
  EXPECT_NE(airLoopHVAC.sizingSystem().handle(), sizingSystem2.handle());  // sometimes true, sometimes not

Expected Behavior

I'm not really sure what should happen. But I do know that results should be deterministic (which they aren't).

Possible Solution

Issue 1: don't allow cloning SizingSystem.
Issue 2: A) don't make the SizingSystem ctor public, or B) ctor call removes existing SizingSystem.

Start by locating the model tests for SizingSystem and read the implementations of AirLoopHVAC.sizingSystem(), SizingSystem::clone(), and the SizingSystem(Model, AirLoopHVAC) constructor. Reproduce both examples and determine the intended ownership and replacement behavior. Done means the relevant tests consistently establish deterministic results and document the chosen behavior.