"Using tensorflow backend" even when using pytorch models

[karnawhat]

When importing the profit module, the Using TensorFlow backend. is printed to the console. However, this should not be printed as the default backend should be pytorch. Somewhat related to #47.

Current behavior

"Using TensorFlow backend."

Expected behavior

"Using PyTorch backend."

[karnawhat]

This happens because some submodules of profit use keras to build layers/models, which, in turn, uses the tensorflow to perform computations.

[karnawhat]

Disable the import of keras/tensorflow (in all modules/submodules) if the tensorflow backend is not being used! This will likely be a big commit and will involve changing many imports statements within files. This involves not importing the submodules that requires tensorflow/torch when importing the root modules. To use the backend you want, one has to import the necessary submodule.