toJSON previously converted a tuple into a list and
then converted that list into a vector.
I now run a ST computation that creates a mutable vector of the correct size
and fills it with the right Values.
This improves performance by 45%.
fromJSON previously converted the vector into a list and
then pattern matched that list so that the elements could be parsed.
I now index the vector directly so we don't need to create a list.
This improved performance by 20%.