Quantifying the mini-batching error in Bayesian inference for Adaptive Langevin dynamics