Current and Voltage decay not being learned after `require_grad=true`

[JDRanpariya]

Describe the bug
For CUBA Block Neuron, after setting grad=true fro current and voltage decay, and setting grad=false for synaptic weight, neuron is not learning the expected behavior.

In experiment setup, I'm generating input spike(poisson) and my target spike has a spike whenever input spike has two consecutive spikes, I freeze synaptic weights and setup initial neuron parameters for threshold to 0.5, current decay to 0.4 and voltage decay to 0.7 -> it's able to learn the pattern but when I change current and voltage decay to 0.1, both decay's are learnable but doesn't learn the pattern and they change very slowly, making me think if they are really affected by SpikeTimeLoss.

Screenshots
This is how the Spike Triggered Average and input, target, output spikes look for current and voltage decay of 0.4, 0.7

and this is how it looks for 0.1, 0.1

Environment (please complete the following information):

• Device: Laptop
• OS: Mac
• Lava DL version 0.6