fix: split pipeline into 4 workflows for cross-node scheduling

Multi-workflow with local-path storage + mismatchLabelKeys anti-affinity
forces each workflow onto a different cluster node:
- test: lint + pytest + vitest (node A)
- build-1: scheduler, symbol-registry, ingestion, parser (node B)
- build-2: extractor, aggregation, recommendation, risk (node C)
- build-3: broker-adapter, lake-publisher, query-api, trading-engine, dashboard, superset (node D)
- finalize: integration-test + github mirror (any available node)

.woodpecker/test.yml

@@ -0,0 +1,52 @@
+# Workflow 1: Lint and tests (node A)
+when:
+  event: [push, pull_request]
+
+steps:
+  lint-python:
+    image: registry.celestium.life/dockerhub-cache/library/python:3.12-slim
+    commands:
+      - pip install ruff==0.15.10 --quiet
+      - ruff --version
+      - ruff check --output-format=concise .
+    backend_options:
+      kubernetes:
+        resources:
+          requests:
+            memory: 256Mi
+            cpu: 200m
+          limits:
+            memory: 512Mi
+            cpu: 1000m
+
+  test-python:
+    image: registry.celestium.life/dockerhub-cache/library/python:3.12-slim
+    commands:
+      - pip install -r requirements.txt --quiet
+      - pip install hypothesis --quiet
+      - pytest tests/ -x --tb=short -q --ignore=tests/integration
+    backend_options:
+      kubernetes:
+        resources:
+          requests:
+            memory: 512Mi
+            cpu: 500m
+          limits:
+            memory: 1Gi
+            cpu: 2000m
+
+  test-frontend:
+    image: registry.celestium.life/dockerhub-cache/library/node:24-slim
+    commands:
+      - cd frontend
+      - npm ci
+      - npx vitest --run
+    backend_options:
+      kubernetes:
+        resources:
+          requests:
+            memory: 512Mi
+            cpu: 500m
+          limits:
+            memory: 1Gi
+            cpu: 2000m