More robust ace handling and value modeling #2
Description
As it stands, when we initiate the kernel, we're fed-in a value that's already calculated and static.
If this value consisted of an ace (which provides flexibility in scoring), we don't track that, we just know the current sum.
When aces are added, they're evaluated as 11, and if that causes a bust, they're changed down to 1. However, from there, they don't actually maintain that future flexibility that they would in reality.
We imitate this same logic both in and out of the kernel.
A more robust handling would re-evaluate the proper sum with all present cards each time it's done, including inside the kernel.