Remove no exp usage from logical rule Part II

[NuojCheng]

Description

Followup PR of #3578

This PR deprecates

• activation_kv_batch_no_exp
• activation_q_length_no_exp
• activation_attn_length_no_exp

from logical names.

After this change

• activation_kv_batch always includes "expert" physical axis
• activation_q_length and activation_attn_length does not include "expert"

Other logical names containing "_no_exp" will be deprecated in following PR.

Tests

CI tests protecting attention.py.

Inference test:

NEW_MODEL_DESIGN=1 python src/maxtext/inference/vllm_decode.py src/maxtext/configs/post_train/rl.yml model_name=qwen3-30b-a3b tokenizer_path=Qwen/Qwen3-30B-A3B ici_tensor_parallelism=4 ici_expert_parallelism=1 enable_dp_attention=false hbm_utilization_vllm=0.3 load_parameters_path=gs://parambole-qwen3-moe-verification/unscanned/qwen3-30b-a3b-thinking-2507/14_08_2025/0/items vllm_hf_overrides='{architectures: ["MaxTextForCausalLM"]}' prompt="Suggest some famous landmarks in London." 2>&1 | tee  qwen3_moe_vllm_0.log

Output: https://paste.googleplex.com/6211135167660032

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
• 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, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

[codecov]

Codecov Report

❌ Patch coverage is 75.00000% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/maxtext/layers/attention_op.py	75.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!