Inconsistent behavior for item assignment to vector arrays

Vector Version

1.3.1

Python Version

3.11.4

OS / Environment

Kubuntu Linux 22.04.
vector is installed using pip in a conda environment.

Describe the bug

When trying to assign certain items in a vector numpy array, the behavior is different depending on how it is done.

> v = vector.array({"E": np.arange(4), "px": np.zeros(4), "py": np.zeros(4), "pz": np.zeros(4)})
> print(v)
[(0., 0., 0., 0) (0., 0., 0., 1) (0., 0., 0., 2) (0., 0., 0., 3)]

Assignment with a slice:

> v[1:2] = vector.array({"E": [8], "px": [0], "py": [0], "pz": [0]})
> print(v) # Changes the item correctly
[(0., 0., 0., 0) (0., 0., 0., 8) (0., 0., 0., 2) (0., 0., 0., 3)]

Assignment with a boolean array:

> v[v.E > 2] = vector.array({"E": np.ones(2), "px": np.ones(2), "py": np.ones(2), "pz": np.ones(2)})
> print(v) # Does nothing and fails silently
[(0., 0., 0., 0) (0., 0., 0., 8) (0., 0., 0., 2) (0., 0., 0., 3)]

Assignment with a single index:

> v[1] = vector.array(dict(px=[0.0], py=[0.0], pz=[0.0], E=[3])) # Crashes
TypeError: 'MomentumObject4D' object does not support item assignment

Any additional but relevant log output

No response

Start by reproducing the reported vector.array assignments with slices, boolean arrays, and a single index in Python 3.11. Compare the three indexing behaviors and add regression coverage for the expected consistent assignment behavior; done means the boolean assignment no longer fails silently and single-index assignment has defined behavior.