The previous version of the CTR encoding was incrementing the counter in a weird way, mixing up data from the previous block.
The current fix can correctly decrypt data encoded with AES-CTR using other libraries/languages (currently tested only with Python).
