`wait_for_past_depends_before_skipping` does not seem to have any effect

### Apache Airflow version Airflow 3.1.0 ### What happened? I am part of a data engineering team. We use airflow to orchestrate our data ingestion pipelines. At a high-level, they are made of: - An ingestion task group, that moves data from a source system to our S3 storage - Validation tests - Processing jobs that clean up and prepares the data for use by downstream consumers - Validations tests on the processed data The processing jobs run in Spark. To avoid overhead, I have been looking at using Airflow's skipping mechanisms to skip the rest of the pipeline if there is no data to be ingested. However, we need the processing jobs to run sequentially. For this purpose, we use `depends_on_past`. I then stumbled upon [the PR](https://github.com/apache/airflow/pull/27710) introducing `wait_for_past_depends_before_skipping`, which seems like it would allow me to skip tasks while still preserving the dependencies on past runs. However, the flag does not seem to have an effect. ### What you think should happen instead? The flag should behave as advertised in the base PR and in the docs. ### How to reproduce ```python from airflow import DAG from airflow.models import Param from airflow.providers.standard.operators.empty import EmptyOperator from airflow.providers.standard.operators.python import ShortCircuitOperator from airflow.sdk import get_current_context def check_continue(): ctx = get_current_context() skip_downstream = ctx["params"]["skip"] return not skip_downstream with DAG( dag_id="test_airflow_skip", schedule=None, params={ "skip": Param(default=False, type="boolean"), }, ) as dag: depends_on_past = EmptyOperator( task_id="depends_on_past", depends_on_past=True, wait_for_past_depends_before_skipping=True, ) skip = ShortCircuitOperator( task_id="skip", python_callable=check_continue, ) skip >> depends_on_past ``` 1. Launch a run 2. Then manually set the state of the …