Add SWE-Bench evaluation results

This PR ensures the model shows up at https://huggingface.co/datasets/SWE-bench/SWE-bench_Verified.

This is based on a new feature on the hub: https://huggingface.co/docs/hub/eval-results.

.eval_results/swe_bench_verified.yaml

@@ -0,0 +1,9 @@
+- dataset:
+    id: SWE-bench/SWE-bench_Verified
+    task_id: swe_bench_%_resolved
+  value: 66.0
+  source:
+    url: https://huggingface.co/GAIR/OpenSWE-72B
+    name: Model card
+    user: nielsr
+  notes: SWE-Agent harness