Fix Compiler dashboard bug for fetching (#7419)

# Overview
Fix Compiler Dashboard bug for fetching unnecessary metrics.query

# Issue
The new compiler inductor also rendering and fetches data which does not
provide benchmark results, The metrics we are interested in are ends
with output: 'performance.csv','accuracy.csv' and 'smoketest.csv'. This
filter out the noisy metadata backend metrics during data

# Demo
 Currently in hud, we have backend that has .csv:

[link1](https://hud.pytorch.org/benchmark/compilers_regression?renderGroupId=main&time.start=2025-10-22T00%3A00%3A00.000Z&time.end=2025-10-29T23%3A59%3A59.999Z&filters.repo=pytorch%2Fpytorch&filters.benchmarkName=compiler&filters.backend=&filters.mode=inference&filters.dtype=bfloat16&filters.deviceName=cuda+%28a100%29&filters.device=cuda&filters.arch=a100&lbranch=main&rbranch=main&rcommit.commit=35f3572fa483a8edb101d5765564e1ae274f3d45&rcommit.workflow_id=18915706635&rcommit.date=2025-10-29T19%3A00%3A00Z&rcommit.branch=main&lcommit.commit=ad4dc52bf6bc09fd3680bcb9bc957203c9cb54f5&lcommit.workflow_id=18697188028&lcommit.date=2025-10-22T01%3A00%3A00Z&lcommit.branch=main&maxSampling=110
)

This pr's demo, no such backend exists anymore

[link2](https://torchci-ehh23x5s6-fbopensource.vercel.app/benchmark/compilers_regression?renderGroupId=main&time.start=2025-10-22T00%3A00%3A00.000Z&time.end=2025-10-29T23%3A59%3A59.999Z&filters.repo=pytorch%2Fpytorch&filters.benchmarkName=compiler&filters.backend=&filters.mode=inference&filters.dtype=bfloat16&filters.deviceName=cuda+%28a100%29&filters.device=cuda&filters.arch=a100&lbranch=main&rbranch=main&rcommit.commit=1b655a87ef137d2cc9603a982532c5e033432daa&rcommit.workflow_id=18899846384&rcommit.date=2025-10-29T09%3A00%3A00Z&rcommit.branch=main&lcommit.commit=d01f15152cdf9a4b693d5c768cef31a0b2a5b012&lcommit.workflow_id=18708183777&lcommit.date=2025-10-22T09%3A00%3A00Z&lcommit.branch=main&maxSampling=110)


# Use endWith
we use hardcoded endWith to filter out the data, endsWith() avoids regex
parsing and string scanning, this is much faster than Like %

# Investigation
To reproduce the data that are filtered ouit, use sql below in
clickhouse
```
SELECT DISTINCT benchmark_extra_info['output']
FROM benchmark.oss_ci_benchmark_torchinductor
WHERE
    timestamp >= now() - INTERVAL 2 MONTH
    AND NOT (
        endsWith(benchmark_extra_info['output'], 'performance.csv')
        OR endsWith(benchmark_extra_info['output'], 'accuracy.csv')
        OR endsWith(benchmark_extra_info['output'], 'smoketest.csv')
    )
ORDER BY benchmark_extra_info['output']
```

result we try to filter out
```
/var/lib/jenkins/pytorch/test/test-reports/inference_huggingface.csv
/var/lib/jenkins/pytorch/test/test-reports/inference_timm_models.csv
/var/lib/jenkins/pytorch/test/test-reports/inference_torchbench.csv
/var/lib/jenkins/pytorch/test/test-reports/training_huggingface.csv
/var/lib/jenkins/pytorch/test/test-reports/training_timm_models.csv
/var/lib/jenkins/pytorch/test/test-reports/training_torchbench.csv
/var/lib/jenkins/workspace/test/test-reports/inference_huggingface.csv
/var/lib/jenkins/workspace/test/test-reports/inference_timm_models.csv
/var/lib/jenkins/workspace/test/test-reports/inference_torchbench.csv
/var/lib/jenkins/workspace/test/test-reports/training_huggingface.csv
/var/lib/jenkins/workspace/test/test-reports/training_timm_models.csv
/var/lib/jenkins/workspace/test/test-reports/training_torchbench.csv
```

Author: yangw-dev

torchci/clickhouse_queries/compilers_benchmark_api_commit_query/query.sql

@@ -15,6 +15,11 @@ WHERE
         )
         OR empty({branches: Array(String)})
     )
+    AND NOT (
+        endsWith(benchmark_extra_info['output'], 'huggingface.csv')
+        OR endsWith(benchmark_extra_info['output'], 'torchbench.csv')
+        OR endsWith(benchmark_extra_info['output'], 'timm_models.csv')
+    )
     AND (
         has({suites: Array(String)}, suite)
         OR empty({suites: Array(String)})

torchci/clickhouse_queries/compilers_benchmark_api_query/query.sql

@@ -19,6 +19,11 @@ SELECT
 FROM benchmark.oss_ci_benchmark_torchinductor
 WHERE
     workflow_id IN ({workflows: Array(UInt64)})
+    AND NOT (
+        endsWith(benchmark_extra_info['output'], 'huggingface.csv')
+        OR endsWith(benchmark_extra_info['output'], 'torchbench.csv')
+        OR endsWith(benchmark_extra_info['output'], 'timm_models.csv')
+    )
     AND (
         has(
             {branches: Array(String)},