Fix post-training Docker build for new vllm commit

[khatwanimohit]

Description

Summary

• GCC version: Upgrades TPU Dockerfile base image from python:3.12-slim-bullseye
  (GCC 10.2.1) to python:3.12-slim-bookworm (GCC 12.2.0). The new vllm commit
  (a30addc7) introduced a CMake check requiring GCC ≥ 11.3 for C++20 support
  vllm PR #44923.

• Missing fastapi extras: Restores fastapi>=0.136.3, fastapi-cli,
  fastapi-cloud-cli, and fastar in the generated requirements. The previous
  generate_requirements.sh run resolved fastapi to 0.115.0 (lowest satisfying
  vllm>=0.115.0 + tpu-inference<0.137.0), which predates the fastar/fastapi-cli
  extras added in 0.136.x. Also pins fastapi[standard]>=0.136.3 and fastar in
  base requirements to prevent future regenerations from dropping these.

Notice 1: Once all tests pass, the "pull ready" label will automatically be assigned.
This label is used for administrative purposes. Please do not add it manually.

Notice 2: For external contributions, our settings currently require an approval from a MaxText maintainer to trigger CI tests.

Tests

build_maxtext_docker_image WORKFLOW=post-training completes successfully

Checklist

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

• I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!