Active scalars regression

Describe the bug, what’s wrong, and what you expect:

It’s my opinion that the following is a regression

#1593 introduced a regression where a mesh can have arrays but no active scalars

Take a look at my comment: https://github.com/pyvista/pyvista/pull/1881#issuecomment-981981603

---

To Reproduce

>>> from pyvista import examples
>>> points = examples.download_sparse_points()
>>> points
PolyData (0x7fbc79419fa0)
  N Cells:	1
  N Points:	16
  X Bounds:	-1.500e+01, 1.500e+01
  Y Bounds:	5.000e+00, 3.500e+01
  Z Bounds:	2.327e+00, 1.102e+01
  N Arrays:	1

>>> points.threshold()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/bane/Software/pyvista/pyvista/pyvista/core/filters/data_set.py", line 842, in threshold
    raise ValueError('No arrays present to threshold.')
ValueError: No arrays present to threshold.
>>>

Even though there is 1 array.

---

System Information: on main and we’re seeing it in the docs

[MatthewFlamm]

.threshold() only uses scalars while .streamlines() only uses vectors. If we use this logic for threshold to motivate setting a default scalars, why not for vectors based on streamlines and similar?

Maybe the answer is that this should apply to both scalars or vectors? I’m not aware of any filters that use the other attributes, but I’m not really sure. This could backfire later if any exist or get added later.

It also gets tricky if this will also apply to vectors as some 3D data is actually not a vector but a 3 component scalar. How do we distinguish between them to set a default vectors? I suppose the worst that happens is we plot based on an incorrect type.

[adeak]

We can fix part of this. We now have machinery that can add data arrays both setting them active and not. It seems that (even if we don’t implement auto-activation of unambiguous arrays) all the meshes created by us (in this case in download_sparse_points()) should have their scalars activated.

And a mostly side note: I wonder if the design choice of always activating one interferes with the concept of active scalars/t_coords/normals/etc. which was also recently added (probably in the same PR, in fact).