Support Kedro 0.19

[sfczekalski]

No description provided.

[sfczekalski]

I don't see an option to comment on the code that hasn't been changed, so I'm writing it down here:

I have doubts how the test_runner_fills_missing_datasets test should work - the nodes are run one by one as independent pipelines, and the only dataset which is added to the catalog is the input_data.

When the first node runs, it loads this dataset, but it doesn't save any output.

The second node should read the first node's output, i2, the input dataset gets created, but there is no data to load, as the fist node hasn't saved any result.

I thought that maybe we should add the output datasets of each node to unsatisfied list in the AzurePipelinesRunner, so that the first node will save its output to i2, but in the test, we also check the free results of the third node run, which is empty, if we add its output to the catalog.

[sfczekalski]

Not sure why changing the directory used to be there

[astrojuanlu]

@ElenaKhaustova what would be the best way to do this with the new catalog?