[Bug]: Flink Portable Runner does not respect log level override pipeline option

### What happened?

When investigating #26301, the raw log (stderr output) is very large. Adding pipeline option
`'--sdkHarnessLogLevelOverrides={"org.apache.flink.runtime":"WARN","org.apache.beam.runners.flink.FlinkPipelineRunner":"WARN"}'`
to here: https://github.com/apache/beam/blob/80f1f6c7f6003ed4daed7492489cf382750ce6f0/runners/flink/job-server/flink_job_server.gradle#L132 does not help.

The log level override is set in BeamFnLoggingClient:

And in FlinkPipelineRunner (this affects pipeline submission logs)

The log level is set in BeamFnLoggingClient constructor: https://github.com/apache/beam/blob/80f1f6c7f6003ed4daed7492489cf382750ce6f0/sdks/java/harness/src/main/java/org/apache/beam/fn/harness/logging/BeamFnLoggingClient.java#L115

However this is executed after flink.runtime log flood. It also does not prevent org.apache.beam.runners.flink.FlinkPipelineRunner log flood after pipeline run.

FlinkPipelineRunner.run() also invokes `SdkHarnessOptions.getConfiguredLoggerFromOptions`. This was effective to run non-portable pipelines (#23631) but appearently not effective for portable flink runner. Although I checked that `FlinkPipelineRunner.run()` is executed.

### Issue Priority

Priority: 2 (default / most bugs should be filed as P2)

### Issue Components

- [ ] Component: Python SDK
- [X] Component: Java SDK
- [ ] Component: Go SDK
- [ ] Component: Typescript SDK
- [ ] Component: IO connector
- [ ] Component: Beam examples
- [ ] Component: Beam playground
- [ ] Component: Beam katas
- [ ] Component: Website
- [ ] Component: Spark Runner
- [X] Component: Flink Runner
- [ ] Component: Samza Runner
- [ ] Component: Twister2 Runner
- [ ] Component: Hazelcast Jet Runner
- [ ] Component: Google Cloud Dataflow Runner