fix(k8s/dev): raise dev API limits to 1Gi/1 for image serving

The atlas/grid-square image endpoints decode 31MB+ MRC/TIFF files to PNG;
under a grid/atlas page's request fan-out this overruns the 256Mi/200m dev
pod (OOMKill + ~15s encodes). Raise the development smartem-http-api limits
to 1Gi/1 so local image serving stays stable across a `dev-k8s.sh`
redeploy, instead of relying on a manual post-deploy
`kubectl set resources` patch. Requests are unchanged (Burstable). Dev
environment only; stage/prod sizing is untouched.

Author: vredchenko

k8s/environments/development/smartem-http-api.yaml

@@ -55,9 +55,12 @@ spec:
           requests:
             memory: "128Mi"
             cpu: "100m"
+          # Image endpoints decode 31MB+ MRC/TIFF files to PNG; under a grid/atlas page's
+          # request fan-out this overruns a 256Mi/200m pod (OOMKill + slow encodes). Give the
+          # dev API headroom so local image-serving stays stable.
           limits:
-            memory: "256Mi"
-            cpu: "200m"
+            memory: "1Gi"
+            cpu: "1"
 
 ---