I have a scenario where I have a series of jobs that are triggered in ADF, the jobs are not linked as such but the resulting temporally tables from each job takes up memory of the databricks cluster. If I can clear the notebook state, that would fre...