v1.0.3

• Ability to validate and test with multiple samples per entry (averaging results)
• Max Subsampling layer in CUDA backend (2D only)
• Flipping image option added to the toolset
• Additional constructor with fixed seed for random generator
• preparing_data command split into preparing_training_data and preparing_testing_data
• A couple of minor bug-fixes