Add OpenLineage parent info to EMR Spark steps

[mobuchowski]

EmrAddStepsOperator can now pass OpenLineage parent job metadata to classic EMR Spark steps. The new openlineage_inject_parent_job_info option defaults to openlineage.spark_inject_parent_job_info, matching the other Spark-based Amazon operators.

The injection reuses OpenLineage's Spark property helper and translates its output into tokenized --conf key=value arguments for command-runner.jar steps using spark-submit or run-example. It leaves non-Spark steps, custom Jar steps, and steps with existing parent settings unchanged, and does not mutate the operator's original step configuration.

This lets OpenLineage events emitted by Spark link back to the originating Airflow task run when parent information injection is enabled.

Prior art

This follows the same opt-in injection pattern already used by:

• SparkSubmitOperator in #47508, which introduced generic Spark parent configuration injection.
• GlueJobOperator in #64513, which translates the properties into Glue's Spark --conf format.
• EmrServerlessStartJobOperator in #64807, which injects parent and transport properties into spark-defaults.
• EmrContainerOperator in #66816, which applies the EMR Serverless pattern to EMR on EKS.

Documentation

The general feature is described in the OpenLineage Spark automatic injection documentation.

This PR also updates:

• The OpenLineage Spark guide with all supported EMR operators, classic EMR argument behavior, and an EmrAddStepsOperator example.
• The Amazon EMR operator guide with setup and behavior details.

---

Was generative AI tooling used to co-author this PR?

• Yes — Codex (GPT-5.6)

Generated-by: Codex (GPT-5.6) following the guidelines