[TPU][CI] Update torchxla version in requirement-tpu.txt

[lsy323]

This is a follow up PR of #12334.

In the current CI setup, we use a nightly torch_xla docker image on a specific day as the base image, then run pip install -r requirements-tpu.txt to install all the TPU dependency packages. However, requirements-tpu.txt includes the nightly torch_xla pinned on a specific day, and it will overwrite the one in the nightly container.

In #12334, the torch_xla version in requirements-tpu.txt wasn't updated so that CI is still using the old torch_xla nightly. In this PR, the nightly version in requirements-tpu.txt is updated.

Maybe it's redundant to use a nightly docker on a specific date as the base image, because the nightly torch_xla in the docker image will be overwritten anyway. This should unblock #12294.

Next step:

• Use a stable base image for TPU CI, instead of a nightly docker. Or other potential way to avoid confusion.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[lsy323]

cc @robertgshaw2-redhat @mgoin @WoosukKwon

[bvrockwell]

Thanks @lsy323 ! Going to just loop in @richardsliu who worked on the original PR that cleaned up a lot of the req file and docs.