Support string columns in load_csv

[jl-wynen]

We are not handling strings well.
Given a file like

idx,name
1,a
2,b
3,c

we get

>>> sc.io.load_csv('asd.csv')
<scipp.Dataset>
Dimensions: Sizes[row:3, ]
Data:
  idx                         int64  [dimensionless]  (row)  [1, 2, 3]
  name                     PyObject        <no unit>  (row)  [a, b, c]

The problem is that the dtype of name is PyObject. I think this happens because pandas defaults to an object dtype. If it used StringDtype, this might work. See https://pandas.pydata.org/pandas-docs/stable/user_guide/text.html

[YooSunYoung]

I also ran into this problem when I was using from_pandas...
I tried to set the dtype as string if all elements in the column are string.
But I'm not sure what should be done when there are int and string.
Should it just fall back to PyObject...?

[YooSunYoung]

Or... can we enable astype from PyObject to string if possible...?