bug distributed array from scalar

[mtar]

Description
factories.array does not support 0-dim tensors if split is not None.

To Reproduce
Steps to reproduce the behavior:

1. Which module/class/function is affected?
   factories.array / stride_tricks.sanitize_axis
2. What are the circumstances under which the bug appears?
   create splitted heat array from scalar
3. What is the exact error-message/erroneous behaviour?

Traceback (most recent call last):
  File "pt.py", line 4, in <module>
    a = ht.array(1, split=0)
  File "heat/core/factories.py", line 323, in array
    split = sanitize_axis(obj.shape, split)
  File "heat/core/stride_tricks.py", line 109, in sanitize_axis
    raise ValueError("axis {} is out of bounds for shape {}".format(axis, shape))
ValueError: axis 0 is out of bounds for shape torch.Size([])
Traceback (most recent call last):
  File "pt.py", line 4, in <module>
    a = ht.array(1, split=0)
  File "heat/core/factories.py", line 323, in array
    split = sanitize_axis(obj.shape, split)
  File "heat/core/stride_tricks.py", line 109, in sanitize_axis
    raise ValueError("axis {} is out of bounds for shape {}".format(axis, shape))
ValueError: axis 0 is out of bounds for shape torch.Size([])

Expected behaviour
Create 0-dim heat tensor

Illustrative

a = ht.array(1, split=0)

b = ht.array(torch.tensor(0), split=0)

Version Info
v0.3.0

Additional comments
This bug affects issue #490

[ClaudiaComito]

@mtar do you think this is causing #490? I don't see why the tests wouldn't fail with 2 processes already if this were the problem.

(why distribute a scalar would be my next question)

[mtar]

The bug happens more specifically in indexing.nonzero() lines 74-75 when lcl_nonzero has only one element:

if a.numdims == 1:
    lcl_nonzero = lcl_nonzero.squeeze()
return factories.array(lcl_nonzero, is_split=is_split, device=a.device, comm=a.comm)

In this particular case, squeeze will return a 0-dim tensor and stuff happens.

On seven processes, you run into it (line 346).