Enable XLA_USE_BF16 for hf-llm tests

[bhavya01]

Description

Please include a summary of relevant context/issue and your changes.

To fix the error INTERNAL: during context [pre-optimization]: Seen floating point types of different precisions in %add.832 = f32[8192,3072]{1,0} add(f32[8192,3072]{1,0} %dot.827, bf16[8192,3072]{1,0} %broadcast.831), but mixed precision is disallowed. while running hf-llm test.

Tests

Please describe the tests that you ran on TPUs to verify changes.

Instruction and/or command lines to reproduce your tests:

./scripts/format-jsonnet.sh
./scripts/gen-tests.sh
./scripts/run-oneshot.sh -t pt-2.2-hf-gpt2-2b-pjrt-conv-v4-8-1vm

List links for your tests (use go/shortn-gen for any internal link):
http://shortn/_H3YSIzI3oi

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.

[yeounoh]

LGTM

[yeounoh]

Thank you @bhavya01