[BUG] Less/greater than binops with `NA` should return `False` in pandas.compatible mode

[mroeschke]

Steps/Code to reproduce bug

In [21]: import pandas as pd

In [22]: import cudf

In [23]: pd_ser = pd.Series([np.nan, np.inf, 2.0])

In [24]: pd_ser < 2
Out[24]: 
0    False
1    False
2    False
dtype: bool

In [25]: cudf.Series.from_pandas(pd_ser, nan_as_null=True) < 2
Out[25]: 
0     <NA>
1    False
2    False
dtype: bool

For cudf.pandas we should be returning False instead of propagating NA

Addressing this will allow us to unskip tests/expr_and_series/is_finite_test.py::test_is_finite_expr[pandas] and tests/expr_and_series/is_finite_test.py::test_is_finite_series[pandas] in #18248