Sparkmeasure

- new: aggregate_stagemetrics and aggregate_taskmetrics added to the Python API
- added Scala and Python tests
- improvements to the documentation
- improvements to the GitHub workflows

Pushed new release to fix issues with 0.21 not being correctly published, see 43

Added print memory report to stage-level metrics collection.
Example:

(scala)> stageMetrics.printMemoryReport
(python)> stagemetrics.print_memory_report()

- The Apache Spark dependency has been bumped up to Spark 3.3.0
- This release is cross compiled for Scala 2.12 and 2.13

Refactoring of StageMetrics and TaskMetrics:
- report will not use Spark SQL but rather compute the metrics using Scala code.
- removed API for accumulables metrics capture and reporting
- added a short report "Stages and their duration" to StageMetrics report
- updated the Python package sparkmeasure match the Scala code refactoring

Improved documentation:
- notably added documentation for the KafkaSink and a reference guide of sparkMeasure API

Latest release to build with Spark 2.4.8 and Scala 2.11
- use this release if you run sparkMeasure with Spark 2.x and Scala 2.11

This adds the Apache Kafka sink.
The versions of several dependencies have been bumped up.