UnPackTo disable merge by default

[dbaileychess]

Fixes: #7458

Adds a new field bool _merge = false to UnPackTo that dictates if the receiving object's existing state is merged with the incoming state or not. When _merge == false, it will reset any pointer in the receiving object to nullptr if there is nothing to unpack. For vectors, the vector is cleared (resized to 0) and then shrunk to reclaim memory.

Previously, the merging of states was on by default, but I don't think that is the correct default value. See #7458 for background discussion. So this change makes merging off by default, and have to be explicitly opted in.

This is tested in two new tests that test with and without merging.

[dbaileychess]

@RWauson FYI

[aardappel]

I think it is admirable that you are supporting both the "merge" and "reuse" use cases, but like I said in the linked issue, "reuse" was the intended use case from the start, and "merge" is a unintended niche use case.

Since supporting code complicates the code (generates more, and slower code), I personally would vote for only supporting reuse. But up to you.

[dbaileychess]

My only reason to include it as an option is because I am switching defaults (I don't know when it was originally switched away from what we wanted). So having the parameter gives people an easy way to go back to the old way.

[aardappel]

That old way was never intended though, the amount of people relying on that behavior should be minimal. I'd be fine not supporting that.

[aardappel]

do you feel the shrink_to_fit is worth it? likely if something is reused, it will have a similar amount of elements, and shrink_to_fit might entail alloc/copy/dealloc?

[dbaileychess]

Yeah, I can remove the shrink to fit part.

I also removed the merge parameter, so it is non-mergeable by default.