Text this: Kernel Reinforcement Learning-Assisted Adaptive Decoder Facilitates Stable and Continuous Brain Control Tasks