Allow Builtin Transforms to be saved (without expansion) in CTF

[doug-walker]

Currently, Builtin Transforms are expanded into the underlying op pieces before being saved to CTF. Although this is exactly equivalent in terms of the numerical results, it could obfuscate the original contents of a GroupTransform.

For example, in the ACES Config generator, it would be convenient to be able to write a Builtin to the CTF, so that the generator could more simply create a Builtin entry in the config's Yaml. (A specific case is with the Canon Log, which becomes a Lut1D when the Builtin is expanded.)

However, in other cases, people may prefer the current behavior. For example, when debugging processing issues, one wants to see the actual ops that a Builtin Transform would be expanded into. So ideally there should be a way to obtain either behavior.

[zachlewis]

This is a great idea. Maybe we could use an optimization flag for governing expansion behavior?

If one wants to see the actual ops, it shouldn't be too difficult to derive the expanded ops from the collapsed form with the API.