🐛 fix: handle edge cases in _compute_progress_from_tasks and clean up query_run_costs

- Use `run.get("tasks") or []` to handle explicit null tasks from JSON
- Use `is not None` check for peakRss to avoid treating 0 as missing
- Fix pcpu comment to correctly describe aggregate CPU% semantics
- Revert query_run_costs to two separate clean SQL queries matching file convention
- Fix docstring to reference Seqera Platform tw run dumps instead of Nextflow log tarballs

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: FloWuenne

bin/benchmark_report.py

@@ -61,10 +61,10 @@ def _task_payload(task_raw: dict) -> dict:
 def _compute_progress_from_tasks(run: dict) -> dict:
     """Compute workflowProgress metrics from task-level data.
 
-    This is a fallback for runs where the Platform did not provide
-    aggregate progress (e.g. imported from Nextflow log tarballs).
+    This is a fallback for runs where aggregate progress is missing
+    (e.g. imported from Seqera Platform tw run dumps).
     """
-    tasks = [_task_payload(t) for t in run.get("tasks", [])]
+    tasks = [_task_payload(t) for t in run.get("tasks") or []]
     completed = [t for t in tasks if t.get("status") == "COMPLETED"]
     if not completed:
         return {}
@@ -77,8 +77,9 @@ def _val(d: dict, key: str, default: int = 0) -> int | float:
     cpu_time = sum(
         _val(t, "cpus") * _val(t, "realtime") for t in completed
     )
-    # cpuLoad = actual CPU usage: pcpu is % of a single core,
-    # so pcpu/100 * realtime gives core-milliseconds used.
+    # cpuLoad = actual CPU usage: pcpu is aggregate CPU% across all cores
+    # (e.g. 400 for a 4-core task at full load), so pcpu/100 * realtime
+    # gives core-milliseconds used.
     cpu_load = sum(
         _val(t, "pcpu") / 100.0 * _val(t, "realtime")
         for t in completed