Add backend interface

[nastiapetrovych]

Add:

• Added backend interface for Torch and TF backends
• Moved quantizers to src/pquant/core/activations_quantizer.py
• Moved compressed layers to src/pquant/core/compressed_layers.py
• Fixed a minor bug layer.name + "_pointwise" in get_enable_pruning() in src/pquant/core/tf_backend.py

To Do:

• Test modifications
• Come up with an idea how to distinguish QuantizedPooling for Keras and Torch (now they are in seperate modules)
• Rewrite test modules (as they use the previous backends functions)