Add always rounding up functionality

[madhur-tandon]

For float values of the type x.5, this round function will always round up to (x+1). Works for both scalar values and N-dimensional numpy arrays / matrices.