Feat: Add MaxText GRPO training DAG for RL pipeline validation

[RexBearIU]

Description

This pull request introduces a new GRPO (Group Relative Policy Optimization) training DAG for the MaxText reinforcement learning pipeline, specifically targeting the Llama3.1 70B model. It also updates workflow triggers and adds support for a new nightly GRPO Docker image.

Airflow/Composer

• GCP Composer name: jackyf-test (under GCP project: cloud-ml-auto-solutions)
• GCP Composer version: 2.13.1

List links for your tests (use go/shortn-gen for any internal link):

jacky-test (Success) https://06ba93284e31466eb62067a2f46710a7-dot-us-east1.composer.googleusercontent.com/dags/maxtext_grpo_rl/grid?dag_run_id=manual__2025-10-03T08%3A27%3A25.167435%2B00%3A00

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run one-shot tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed.