Filtering by a None value in a string column doesn't work.

[mdymczyk]

When there's a missing value in the column, if I get all the unique values from it using dt.unique I can then filter rows using frame[f.column == uniques[idx]] for every unique value but the missing None one, which fails with an error. Now whether this is working as intended or not I'm not sure but at least I think we can improve the error message.

Repro:

>>> import datatable as dt
>>> from datatable import f
>>> dataset = dt.Frame([[None, '1','2']], names=['a'])
>>> ref_levels = dt.unique(dataset[:, 'a']).to_list()[0]
>>> dataset[f.a == ref_levels[2], :]
    a
--  --
 0  2

[1 row x 1 column]
>>> dataset[f.a == ref_levels[1], :]
    a
--  --
 0  1

[1 row x 1 column]
>>> dataset[f.a == ref_levels[0], :]
TypeError: Binary operator == cannot be applied to columns with stypes str32 and bool8

>>> ref_levels
[None, '1', '2']

I tried the above with some numeric types (also floats) and all seems to work fine then.