DeepDecipher: Accessing and Investigating Neuron Activation in Large Language Models