v0.0.1 Release

v0.0.1 Release.
Improved gradient computing.
Now the library supports lazy backward pass, preventing recomputing the same pathes many times, and eventually removing exponential growth of the computational cost of LSTM or similar arcitecures.

Full Changelog: https://github.com/Hacker1337/Fast-RNN-DL-Framework/commits/v0.0.1