Card detection revamping

[Masuzu]

Machine learning for card value identification (2, 3, 4,..., K, A, Joker). Successfully tested on medium/large browser sizes with low/high resolution settings. Recalibrations are no longer needed with the new system and 100% accuracy for card value identification \o/ The system is robust to changes in browser window zoom and/or Window scaling settings, but not yet fully implemented. Tests made on an i7 4790 and GTX 980Ti

Pitfalls:

• card suit (Spade, Heard, Clubs, Diamond) identification may not be 100% accurate (yet)
• the new archive size is much larger than before. A downloader will be added to retrieve the required dependency automagically.
• potential hardware compability issues for people not having CUDA GPUs (to be checked)
• the fallback to CPU-only computations instead of using the GPU is very likely induce a loss in evaluation speed which needs to be tested
• will be made available for 64-bit OS only

Card suit identification: work in progress.

[Masuzu]

New version 2.0 with pattern recognition for card detection now live: https://poker-bot.azurewebsites.net/en/Home/PatchNotes