Support for mixed precision in JACC.parallel_reduce

[tdehoff]

I was using JACC.parallel_reduce to perform reductions on arrays of smaller integer types, such as Int8. In many cases, the result of such a reduction can overflow, so I need the accumulator to be a wider type, such as Int64.
Is it possible to support mixed-precision reductions, where the input array is of one type and the result is another?

Thanks!