Segfault when creating frames from `np.array()` of unicode strings

[Oleksiy]

import datatable as dt
import numpy as np
src = ['ы']*100
DT = dt.Frame(np.array(src)) # either a segfault, or the last few rows contain corrupted data

At the same time, creating a frame from a python list of unicode strings works just fine, i.e.

DT = dt.Frame(src) # no segfault, frame contains valid data

Originally posted by @arnocandel