ci: fix airflow timeout flakiness with 4 mitigations

1. Reduce airflow workers from 2 to 1 (less k8s resource contention)
2. Increase minikube memory from 6GB to 7GB for airflow backend
3. Increase per-test timeout from 900s to 1200s for airflow
4. Add --reruns=1 across all backends to retry flaky timeouts

The airflow deployer tests have high overhead per test (DAG parsing,
k8s pod scheduling, conda env setup) that sometimes exceeds 900s in
constrained CI environments. Different tests timeout on each run,
confirming this is environmental variance not a code bug.

Author: Nissan Pow

.github/workflows/ux-tests.yml

@@ -100,15 +100,23 @@ jobs:
           - backend: local
             services: "minio,postgresql,metadata-service"
             workers: 4
+            memory: 6144
+            timeout: 900
           - backend: argo-kubernetes
             services: "minio,postgresql,metadata-service,argo-workflows"
             workers: 2
+            memory: 6144
+            timeout: 900
           - backend: airflow-kubernetes
             services: "minio,postgresql,metadata-service,airflow"
-            workers: 2
+            workers: 1
+            memory: 7168
+            timeout: 1200
           - backend: sfn-batch
             services: "minio,postgresql,metadata-service,localbatch,ddb-local,sfn-local"
             workers: 2
+            memory: 6144
+            timeout: 900
 
     runs-on: ubuntu-latest
 
@@ -138,7 +146,7 @@ jobs:
         with:
           driver: docker
           cpus: 2
-          memory: 6144
+          memory: ${{ matrix.memory }}
 
       - name: Restore minikube image cache
         id: image-cache
@@ -259,7 +267,8 @@ jobs:
             -n ${{ matrix.workers }} \
             -v \
             --tb=short \
-            --timeout=900 \
+            --timeout=${{ matrix.timeout }} \
+            --reruns=1 \
             --cov=metaflow \
             --cov-report=term-missing \
             --cov-report=xml:coverage.xml \