Report avg_inference_latency from Android LLM benchmark app

[guangy10]

🚀 The feature, motivation and pitch

Missing from Android. See details here: #8576 (comment)

Alternatives

No response

Additional context

No response

RFC (Optional)

No response

cc @huydhn @kirklandsign @shoumikhin

[kirklandsign]

Sorry just realize for LLM it's missing

[kirklandsign]

Numbers are reported now https://hud.pytorch.org/benchmark/llms?startTime=Wed,%2016%20Apr%202025%2017:26:16%20GMT&stopTime=Wed,%2023%20Apr%202025%2017:26:16%20GMT&granularity=day&repoName=pytorch%2Fexecutorch&benchmarkName=&modelName=llama&backendName=xnnpack_q8&modeName=inference&dtypeName=&deviceName=Samsung%20Galaxy%20S22%205G%20(Android%2013)&archName=All%20Platforms

[kirklandsign]

#10376