GENNI: Visualising the Geometry of Equivalences for Neural Network Identifiability