Version 0.12.6

ADD #886: Provide new function which allows fitting only a single configuration.
DOC #1070: Clarify example on how successive halving and Bayesian optimization play together.
DOC #1112: Fix type.
DOC #1122: Add Python 3 to the installation command for Ubuntu.
FIX #1114: Fix a bug which made printing dummy models fail.
FIX #1117: Fix a bug previously made memory_limit=None fail.
FIX #1121: Fix an edge case which could decrease performance in Auto-sklearn 2.0 when using cross-validation with iterative fitting.
FIX #1123: Fix a bug autosklearn.metrics.calculate_score for metrics/scores which need to be minimized where the function previously returned the loss and not the score.
FIX #1115/#1124: Fix a bug which would prevent Auto-sklearn from computing meta-features in the multiprocessing case.

Contributors v0.12.6