[BUG] loc returning incorrect results for DatetimeIndex that is in monotonically decreasing

[galipremsagar]

Describe the bug
When we perform a sliced loc on a DatetimeIndex that is monotonically decreasing, the result doesn't match with pandas.

Steps/Code to reproduce bug

In [1]: import cudf

In [4]: df = cudf.DataFrame({'a': [1, 1, 1, 2, 2], 'b': [1,2 ,3, 4, 5]}, index=cudf.Index([15, 14, 12, 10, 1], dtype='datetime64[ns]'))

In [5]: df
Out[5]: 
                               a  b
1970-01-01 00:00:00.000000015  1  1
1970-01-01 00:00:00.000000014  1  2
1970-01-01 00:00:00.000000012  1  3
1970-01-01 00:00:00.000000010  2  4
1970-01-01 00:00:00.000000001  2  5


In [9]: import pandas as pd

In [10]: pd.Timestamp("1970-01-01 00:00:00.000000010")
Out[10]: Timestamp('1970-01-01 00:00:00.000000010')

In [11]: i = pd.Timestamp("1970-01-01 00:00:00.000000010")

In [12]: df.loc[i:]
Out[12]: 
                               a  b
1970-01-01 00:00:00.000000015  1  1
1970-01-01 00:00:00.000000014  1  2
1970-01-01 00:00:00.000000012  1  3
1970-01-01 00:00:00.000000010  2  4

In [13]: df.to_pandas().loc[i:]
Out[13]: 
                               a  b
1970-01-01 00:00:00.000000010  2  4
1970-01-01 00:00:00.000000001  2  5

Expected behavior
Match pandas.