Construct ndarray/array from iterator/generator

Reproducing code example:

I was trying to calculate a mean from an array. Then I decided to do some filtering, which changes the input array/list to an iterator.

np.mean([0,1,2]) # OK
np.mean(value for value in [0,1,2]if value > 0) # CRASH

because internally mean calls array(sequence). And this happens:

>>> np.array(iter([1,2]))
array(<list_iterator object at 0x000001CFC0F96608>, dtype=object)

>>> np.ndarray(iter([1,2]))  # for reference
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: expected sequence object with len >= 0 or a single integer

It seems very weird for generators and iterators to just be placed inside an array, and counter-intuitive compared to many other stdlib functions, like all(), any(), sum(), iter(), list(), etc...

Suggestion: traverse iterators and generators when creating ndarray and array and make the array with its elements.
There might be someone that wants to preserve the old behavior, so one could write array([iterator])

Numpy/Python version information:

1.18.1 3.7.5 (tags/v3.7.5:5c02a39a0b, Oct 15 2019, 00:11:34) [MSC v.1916 64 bit (AMD64)]

Start from the np.array and np.ndarray construction entry points demonstrated in the report, and reproduce the iterator and generator examples on the stated NumPy/Python versions. Investigate how iterator inputs should be distinguished from the existing array([iterator]) behavior. Done means the intended construction semantics are decided and covered for both examples.