Countna()

[xdssio]

The idea here is to get a quick missing values count.

I have considered returning a dict, but eventually settled on a pandas series for two main reasons: a. It's similar to value_counts() b. It is very easy to either sum it, or to_dict the result which is the most common cases.

Example

import vaex
import numpy as np
df = vaex.from_arrays(x=[1,2,np.nan], y=['a','b',None])
df.countna()

x    1
y    1
dtype: int64

[JovanVeljanoski]

I am happy with this PR.

Long term we probably do not want to return a pandas object, but since the same goes from things like describe, value_counts, this is probably ok too.

The question is do we want to also do the same for countnan and countmissing ? Or maybe include them on demand?