`v.perp(direction)` should compute 2D magnitude along an alternate 3D longitudinal axis direction (both are at least 3D)

From ROOT's TLorentzVector:

>>> import ROOT

>>> v = ROOT.TLorentzVector(1.1, 2.2, 3.3, 4.4)
>>> direction = ROOT.TVector3(3, 2, 1)

>>> v.Perp(direction)
2.880476151115099

We want the same thing in a convenient method: v.perp(direction)

>>> import awkward as ak
>>> import numpy as np
>>> import vector
>>> vector.register_awkward()

>>> v = ak.Array([
...     [{"x": 1.1, "y": 2.2, "z": 3.3}], [], [{"x": 1.1, "y": 2.2, "z": 3.3}]
... ], with_name="Vector3D")

>>> direction = ak.Array([
...     [{"x": 3, "y": 2, "z": 1}], [], [{"x": 3, "y": 2, "z": 1}]
... ], with_name="Vector3D")

>>> v.mag * np.sin(v.deltaangle(direction))
<Array [[2.88], [], [2.88]] type='3 * var * float64'>

New "compute" modules are not needed; just compute this as a convenience method from the existing ones.

Locate the existing perp, mag, and deltaangle implementations and follow how Vector3D arrays and empty elements are handled. Add the convenience behavior for v.perp(direction) without a new compute module, matching ROOT's result for the shown 3D inputs and preserving the demonstrated awkward-array structure.