[C++] Enable ULP-based float comparison for Arrow Scalar, Array, RecordBatch, Table, Tensor, and SparseTensor

[andishgar]

Describe the enhancement requested

As I suggested here, it would be beneficial to support ULP-distance-based float comparisons for Scalar, Array, RecordBatch, Table, Tensor, and SparseTensor.

Component(s)

C++