fix(orchestration): Fix cancellation propagation in workflow and concurrent executor

Root causes:
1. WorkflowEngine.ExecuteStepAsync was catching OperationCanceledException
   and wrapping it as a failed step result instead of propagating it
2. ConcurrentExecutor had similar issue with exception handling

Changes:
1. Added explicit catch for OperationCanceledException in:
   - WorkflowEngine.ExecuteStepAsync (re-throws immediately)
   - ConcurrentExecutor operation handler (re-throws before error handling)

2. This allows cancellation tokens to work properly in all tests,
   preventing 10+ second delays from cancelled operations

Results:
- WorkflowEngineTests.ExecuteAsync_SupportsCancellation now passes
- Cancellation properly propagates through entire execution chain
- Tests with long delays (10s, 5s, 2s) can now be cancelled immediately

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

Author: Ahmed Mustafa

src/DotNetDevMCP.Orchestration/ConcurrentExecutor.cs

@@ -92,6 +92,11 @@ await Parallel.ForEachAsync(
 
                         ReportProgress(progress, totalOperations, completedCount, failedCount);
                     }
+                    catch (OperationCanceledException)
+                    {
+                        // Always propagate cancellation
+                        throw;
+                    }
                     catch (Exception ex)
                     {
                         var error = new ExecutionError(

src/DotNetDevMCP.Orchestration/WorkflowEngine.cs

@@ -173,6 +173,12 @@ private async Task<StepExecutionResult> ExecuteStepAsync(
                 Error: result.ErrorMessage,
                 Duration: stepwatch.Elapsed);
         }
+        catch (OperationCanceledException)
+        {
+            // Allow cancellation to propagate
+            stepwatch.Stop();
+            throw;
+        }
         catch (Exception ex)
         {
             stepwatch.Stop();