[JAX] Fix MNIST L2 jax test instability

[tdophung]

Description

L2 jax unittest for mnist sometimes fails in CI for non-error reasons such as noisy loss near convergence. This PR fixes that with looser thresholds, and checking the min loss in the last 10% of steps to filter out noisy high loss near convergence. Also apply xla flags for deterministic results

Fixes # (issue)

Type of change

• Documentation change (change only to the documentation, either a fix or a new content)
• 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)
• Infra/Build change
• Code refactoring

Changes

• Loosen loss failure threshold in MNIST test
• Add xla gpu deterministic flag for this test
• Fix typo for test accuracy check (9.8% -> 98%)

Checklist:

• I have read and followed the contributing guidelines
• The functionality is complete
• 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

[tdophung]

/te-ci jax L2

[tdophung]

/te-ci jax L2