Utilize in place algorithms to minimise memory freeing and new memory allocating

[willayy]

Primarily thinking about these lines in the training algorithm:

SNNPy/snn/src/neuralNetworkTraining.c

Lines 57 to 60 in 1f1e30c

	double * partialGradients = (double *) malloc(sizeof(double) * nn->nrOfNeurons);
	for (int i = 0; i < nn->nrOfNeurons; i++) { partialGradients[i] = 0;}
	GradientVector * gv = (GradientVector *) malloc(sizeof(GradientVector));
	initGradientVector(gv, nn->nrOfNeurons);

SNNPy/snn/src/neuralNetworkTraining.c

Lines 124 to 126 in 1f1e30c

	GradientVector * gv = (GradientVector *) malloc(sizeof(GradientVector));
	initGradientVector(gv, gb->gradientVectors[0]->nrOfNeurons);

SNNPy/snn/src/neuralNetworkTraining.c

Lines 132 to 133 in 1f1e30c

	NeuronGradient * ng = (NeuronGradient *) malloc(sizeof(NeuronGradient));
	initGradient(ng, gb->gradientVectors[i]->gradients[j]->nrOfWeights);