[userbenchmark] Broader error catching

- Failures on a recent sample run indicate that errors raised by the
subprocess are not always Exceptions, but are still sometimes
recoverable
- Add `except` clause to catch all such errors, short of keyboard
interrupts, so the compilation can complete despite such errors

userbenchmark/torch_trt/run.py

@@ -271,10 +271,15 @@ def run(args: List[str]):
                 # Delete model instance and clean up workspace
                 del Model
 
-            except Exception as e:
+            # Halt further model runs
+            except KeyboardInterrupt:
+                break
+
+            # Any subprocess errors should be caught and reported
+            except:
                 traceback.print_exc()
                 print(
-                    f"\nLoading model {model_name} failed with:\n{e}\nSkipping the model.\n"
+                    f"\nBenchmarking model {model_name} failed.\nSkipping the model.\n"
                 )
                 metrics = {
                     model_name: f"Failed to run benchmark: {traceback.format_exc()}"