Neuron models allocate too much memory for master population tables

The size for the the MPT memory section is calculated in bytes, while the memory handler uses 32-bit words for specifying sizes.

example from the LIF Neuron implementation: https://gitlab.com/spinnaker2/py-spinnaker2/-/blob/main/src/spinnaker2/neuron_models/lif_neuron.py?ref_type=main#L189

        mpt_n_bytes = mpt_length * N_WORDS_MPT_ENTRY * 4

        memory_handler.add_mem_region(
            name=self.regions.MPT.name, index=self.regions.MPT.value, data=None, size=mpt_n_bytes, send=True, read=False
        )

Just removing the * 4 seems to work in my case, but I haven't done any extensive testing.