build: upgrade Flink from 1.18.1 to 1.20.5

[romitshah02]

Upgrades Apache Flink from 1.18.1 to 1.20.5 across all job modules. Migrates off the deprecated flink-streaming-scala API (removed in Flink 2.0) to flink-streaming-java/Java DataStream types, bumps flink-connector-kafka to 3.4.0-1.20, and fixes a stale log4j pin in the Docker image that crashed the TaskManager on startup against the new base image.

Dependencies required: org.apache.flink:flink-connector-kafka:3.4.0-1.20, org.apache.flink:flink-streaming-java:1.20.5 (replacing flink-streaming-scala_2.12), log4j2 2.24.3 (matches what Flink 1.20.5 already bundles).

Type of change

Please choose appropriate options.

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• Any dependent changes have been merged and published in downstream modules

[sntiwari1]

Summary

Clean, well-scoped Flink 1.18.1 → 1.20.5 upgrade. The core migration pattern — dropping flink-streaming-scala (removed in Flink 2.0) for flink-streaming-java/Java DataStream, switching OutputTag construction to the explicit (name, TypeInformation) constructor now that the implicit Scala macro is gone, and widening Iterable[I] → java.lang.Iterable[I] in the WindowBaseProcessFunction/TimeWindowBaseProcessFunction process() signatures — is applied consistently across every job module (lms-jobs, ml-jobs, notification, user-org-jobs). I verified:

• No leftover references to org.apache.flink.streaming.api.scala or flink-streaming-scala anywhere in the tree (git grep clean).
• WindowBaseProcessFunction/TimeWindowBaseProcessFunction have no subclasses in the repo, so the breaking process() signature change is safe.
• The stringTypeInfo/notificationMetaTypeInfo/userFeeMetaTypeInfo implicits referenced by the new OutputTag constructors are already declared in each touched config class, so nothing is missing.
• flink-connector-kafka:3.4.0-1.20 usage in FlinkKafkaConnector already relies on the KafkaSource builder API (not the legacy FlinkKafkaConsumer), so no additional migration is needed there.
• The add-jira-description CI failure on this PR is an unrelated Jira automation check, not a build/test failure — no CI job actually compiles/tests this PR currently.

Only minor, non-blocking observations below — nothing I'd hold the merge for.