[kwang3939]

Description

Fixes three breakages against upstream vLLM HEAD:

1. fp8 MoE AttributeError (DeepSeek torchax E2E): vLLM #57092 reads self.per_act_token_quant in Fp8MoEMethod.get_fused_moe_quant_config. VllmFp8MoEMethod replaces vLLM's __init__, so the attribute was never set. Set it to False; the TPU kernels do not use the resulting quant config.
2. Continue Decode placeholder underflow (test_sched_utils.py): vLLM #61589 moved num_output_placeholders -= N and its assert from AsyncScheduler into the base Scheduler, so it now runs before the patch's post-call compensation. Pre-compensate by N - 1 before calling the base instead, which works on both LKG and HEAD. Also fix the test fixture so it does not pin the unpatched base onto AsyncScheduler.
3. Routed-experts connector tests (test_routed_experts_slots.py): vLLM #62591 changed AuxOutputSchedulerConnector.__init__ to take a KVCacheManager. Pass a minimal stub from the test helper. Test-only.

Tests

Tested with:

python3 -m pytest -s -v tests/layers/vllm/test_fp8.py::test_fused_moe
python3 -m pytest tests/core/test_sched_utils.py tests/runner/test_routed_experts_slots.py

Checklist

Before submitting this PR, please make sure:

• I have performed a self-review of my code.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have made or will make corresponding changes to any relevant documentation.

[github-actions]

Description

Start with a short description of what the PR does and how this is a change from
the past.

The rest of the description includes relevant details and context, examples:

• why is this change being made,
• the problem being solved and any relevant context,
• why this is a good solution,
• some information about the specific implementation,
• shortcomings of the solution and possible future improvements.

If the change fixes a Github issue, please include a link, e.g.,:
FIXES: #123456

Tests

Please describe how you tested this change, and include any instructions and/or
commands to reproduce.

Checklist

Before submitting this PR, please make sure:

• I have performed a self-review of my code.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have made or will make corresponding changes to any relevant documentation.