Add Float IR node class

[naoyam]

Represents the 32-bit floating-point scalar value. Not supported in PyTorch, so can't be used as inputs to fusions

Intended to be used in some of the low-level optimizations like welford vectorization.

Since the code for Float and Double would be mostly identical, I created the FloatingPoint template class and defined Float and Double as:

using Float = FloatingPoint<DataType::Float>;
using Double = FloatingPoint<DataType::Double>;