Add unique() built-in function

The current unique() builtin uses a script unique.dml, which does not optimized for the Spark backend. We should therefore create an optimized implementation of unique.

The new function should behave similarly to that in R: https://www.geeksforgeeks.org/unique-function-in-r/

Discussion of the requirements can be found here: 

https://github.com/apache/systemds/pull/1714

API Design

1. Row aggregation: remove duplicate rows

    * R: unique() removes duplicate rows, eg

        > df <- matrix(rep(1:6,length.out=9),nrow = 3,ncol=3,byrow = T)
        > df
            [,1] [,2] [,3]
        [1,]    1    2    3
        [2,]    4    5    6
        [3,]    1    2    3
        > unique(df)
            [,1] [,2] [,3]
        [1,]    1    2    3
        [2,]    4    5    6

    * SystemDS: the same can be achieved using the unique() sketch like so:

        unique(X, dir="r")

2. Col aggregation: remove duplicate cols

    * R: unique() removes duplicate rows, so we can obtain the desired result using transpose, like so:

        > df <- matrix(rep(1:6,length.out=9),nrow = 3,ncol=3)
        > df
            [,1] [,2] [,3]
        [1,]    1    4    1
        [2,]    2    5    2
        [3,]    3    6    3
        > df_t
            [,1] [,2] [,3]
        [1,]    1    2    3
        [2,]    4    5    6
        [3,]    1    2    3
        > df_t_ = unique(df_t)
        > df_t_
            [,1] [,2] [,3]
        [1,]    1    2    3
        [2,]    4    5    6
        > df_t_t = t(df_t_)
        > df_t_t
            [,1] [,2]
        [1,]    1    4
        [2,]    2    5
        [3,]    3    6

    * SystemDS: the same can be achieved using the unique() sketch like so:

        unique(X, dir="c")

3. RowCol aggregation: return only unique values in given matrix

    * SystemDS

        X = [[1, 1], [2, 2], [3, 3]]
        unique(X) will return X' = [[1], [2], [3]]
    
    * R

    This is similar to how unique() operates on vectors in R:

        > df <- c(1, 1, 2, 2, 3, 3)
        > df
        [1] 1 1 2 2 3 3
        > unique(df)
        [1] 1 2 3

    The difference is that SystemDS' unique() will support the same for not only vectors, but also matrices.