[FEA] Masked UDF Unary ops

[brandon-b-miller]

Is your feature request related to a problem? Please describe.
Masked UDFs were implemented starting with binary operations, meaning we enabled this:

df = cudf.DataFrame({
    'a':[1,2,3],
    'b':[4,5,6],
})
def f(row):
    return row['a'] + row['b']

df.apply(f, axis=1)

Now we need the set of unary operations as well, i.e.

def f(row):
    theta = row['b']
    return math.sin(theta)

Describe the solution you'd like
Implement typing and lowering in numba to handle these cases for cudf MaskedType. These should be fairly simple to just plug in without having to touch the overall UDF pipeline.

Describe alternatives you've considered
There are not really alternatives other than not supporting this capability.

Additional context
N/A