When dividing by random parameter, one obtains fixed, not random values

[heplesser]

Describe the bug

When initializing neuron or synapse properties with a random parameter appearing as a denominator, NEST assigns fixed values.

To Reproduce

1. (Minimal) reproducing example

import nest
n = nest.Create('iaf_psc_alpha', 10, {'V_m': 1 / nest.random.uniform(0, 1)})
print(n.V_m)

Expected behavior
Expected results: 10 different membrane potential values
Observed result: 10 identical membrane potential values

Desktop/Environment (please complete the following information):

• NEST-Version: nest master @ 8dd4b52

Additional context

We have traced the error to the implementation of __rtruediv__() here:

nest-simulator/pynest/nest/lib/hl_api_types.py

Lines 1086 to 1089 in 1ddea6b

	def __rtruediv__(self, lhs):
	rhs_inv = CreateParameter("constant", {"value": 1 / float(self.GetValue())})
	return rhs_inv._binop("mul", lhs)