Investigate decode neuron performance properties and scaling

[arvoelke]

This is motivated in part by the following TODOs:

nengo-loihi/nengo_loihi/decode_neurons.py

Lines 296 to 298 in 1517915

	self.scale = 1.05 * target_point / (self.dt * target_rate)
	# ^ TODO: why does this 1.05 factor help? found it empirically in
	# test_decode_neurons.test_add_inputs

nengo-loihi/nengo_loihi/decode_neurons.py

Lines 264 to 266 in 1517915

	self.scale = 1.08 * target_point / (self.dt * target_rate)
	# ^ TODO: why does this 1.08 factor help? found it empirically in
	# test_decode_neurons.test_add_inputs

More generally, we should look more carefully at profiling decode neurons (formerly called "interneurons") under a wide variety of conditions (frequency, tau, dynamic range, number of neurons), and characterize / identify their dynamical behaviour, so that we may improve these methods and optimize their parameters.