fix: ascend_scheduler adapt to v0.9.0

[zzzzwwjj]

What this PR does / why we need it?

fix ascend_scheduler for v0.9.0

Does this PR introduce any user-facing change?

How was this patch tested?

[wangxiyuan]

need add a test to make sure ascend scheduler always works

[MengqingCao]

plz add pytest -sv tests/scheduler to https://github.com/vllm-project/vllm-ascend/blob/main/.github/workflows/vllm_ascend_test.yaml#L117 and https://github.com/vllm-project/vllm-ascend/blob/main/.github/workflows/vllm_ascend_test.yaml#L132

[whx-sjtu]

vllm-ascend/vllm_ascend/core/scheduler.py

Line 189 in 5270507

assert num_new_tokens == 1

looks like this fix still doesn't support speculative decoding like MTP, I will just test my next PR pr943 which tries to support both MTP and disaggregated-prefill.