[Task]: Improve documentation of PortableRunner for usage with Spark

[mosche]

What needs to happen?

Users are struggling to use the portable runner with Spark. Related Spark documentation should be updated and provide detailed instructions.

While users will probably still prefer the classic Spark runner for most Java pipelines, the Spark portable runner enables cross-language transforms. We can start by providing instructions for plain Java, then add cross-language (Java -> Python).

See:

• https://stackoverflow.com/questions/66320831/its-possible-to-configure-the-beam-portable-runner-with-the-spark-configuration
• https://stackoverflow.com/questions/66498209/how-to-configure-beam-python-sdk-with-spark-in-a-kubernetes-environment

Also related: #21092

Issue Priority

Priority: 2

Issue Component

Component: runner-spark