feat: store raw processor tree JSON and add error categorization fields

Fixes iteration overlay corruption caused by flat storage collapsing
duplicate processorIds across loop iterations.

Server:
- Store raw processor tree as processors_json JSONB on executions table
- Detail endpoint serves from processors_json (faithful tree), falls back
  to flat record reconstruction for older executions
- V10 migration: processors_json, error categorization (errorType,
  errorCategory, rootCauseType, rootCauseMessage), OTel (traceId, spanId),
  circuit breaker (circuitBreakerState, fallbackTriggered), drops
  erroneous splitDepth/loopDepth columns
- Add all new fields through full ingestion/storage/API chain

UI:
- Fix overlay wrapper filtering: check wrapper type before status filter
- Add new fields to schema.d.ts

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

ui/src/api/schema.d.ts

@@ -1626,6 +1626,12 @@ export interface components {
             attributes: {
                 [key: string]: string;
             };
+            errorType?: string;
+            errorCategory?: string;
+            rootCauseType?: string;
+            rootCauseMessage?: string;
+            traceId?: string;
+            spanId?: string;
         };
         ProcessorNode: {
             processorId: string;
@@ -1653,6 +1659,13 @@ export interface components {
                 [key: string]: string;
             };
             resolvedEndpointUri?: string;
+            errorType?: string;
+            errorCategory?: string;
+            rootCauseType?: string;
+            rootCauseMessage?: string;
+            errorHandlerType?: string;
+            circuitBreakerState?: string;
+            fallbackTriggered?: boolean;
             children: components["schemas"]["ProcessorNode"][];
         };
         DiagramLayout: {