Wav2vec-based Detection and Severity Level Classification of Dysarthria from Speech