GCNConv call() implementation

[cedavidyang]

Based on class documentation, GCNCon should compute $$X' = \hat D^{-1/2} \hat A \hat D^{-1/2} X W + b$$

However, the implementation in call computes the following: $$X' = A X W + b$$

based on line 100-103 in spektral/layers/convolutional/gcn_conv.py, copied below

        output = K.dot(x, self.kernel)
        output = ops.modal_dot(a, output)
        if self.use_bias:
            output = K.bias_add(output, self.bias)