Create a CI workflow that build maxtext package and run tests against it

[kanglant]

Create MaxText Package Tests workflow, which builds maxtext package and run cpu/tpu unit/integration tests.

Save the built maxtext distributions as action artifacts. Artifacts are retained within the workflow run for 90 days.

Create a docker image, which has all the system level deps for maxtext and no python libraries pre-installed except pip and uv, for the package testing in a clean python environment.

Test with local configs/assets/test_assets files for now.

Tests

The MaxText Package Tests presubmit passed.

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 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.

[SamuelMarks]

We're not doing the rename yet; soon; but not yet

[bvandermoon]

LGTM, thanks @kanglant. Please try the two disabled tests again with the dependencies from #2301 to see if it helps. If the tests still fail, then they can be disabled like you have in this PR.

For others, Kanglan is adding new Github Actions tests and still keeping the old ones. So we are still covered by that. We will remove the old ones later when we are satisfied with how they are performing like Kanglan mentioned

[NuojCheng]

LGTM

[Rohan-Bierneni]

Do we want to run these tests as part of every unit test even though they are duplicate of the old ones, the difference is setup is using maxtext package? I think we could have them run only on scheduled runs or a nightly run to make sure they work properly so we can save on resources and also on total unit test time? @kanglant @bvandermoon

[bvandermoon]

Do we want to run these tests as part of every unit test even though they are duplicate of the old ones, the difference is setup is using maxtext package? I think we could have them run only on scheduled runs or a nightly run to make sure they work properly so we can save on resources and also on total unit test time? @kanglant @bvandermoon

@Rohan-Bierneni we are planning to make these the main MaxText CI tests. @kanglant is currently keeping the old ones in place until we build confidence in the new ones. There should not be both running for long. Also as FYI, she is not updating the GPU tests for now

[parambole]

Awesome! Thanks for adding this.