BUG: .cast() to array outputs list instead of np.array in Pandas backend

[timothydijamco]

Problem

In the Pandas backend, when casting a column to an array type, a list is created as the output instead of a np.array (which is the new standard in the Pandas backend).

>>> series = t['arr_col'].execute()
>>> type(series[0])
numpy.ndarray
>>> series = t['arr_col'].cast('array<float64>').execute()
>>> type(series[0])
list

Solution

Change the Pandas backend implementation of the cast operation to produce a np.array.

[icexelloss]

LGTM. Minor comment.

[timothydijamco]

@gerrymanoim This PR is breaking cast-to-array on the Dask backend. I've xfailed the test that was failing. #2743 will fix this. Does that sound reasonable to you? I'd also be happy to instead create a working Dask-specific execution function for Cast in the meantime (currently the Dask backend borrows the Pandas backend's Cast execution function).

[jreback]

lgtm @timothydijamco ping when ready

[icexelloss]

LGTM!

[timothydijamco]

thanks, should be all set @jreback

[jreback]

thanks @timothydijamco