Build: Support M1 Processors

Currently the integration tests fail on M1 Processor based machines because the Spark version we specify for running integration tests lacks a Snappy with a compatible native-lib (and proper fallback.)

We can fix this by upgrading to a Spark version which has the updated Snappy.
https://issues.apache.org/jira/browse/SPARK-33725
2.4.8 and 3.0.2 (3.1.0 should already be ok)