fix: update startBlock to current chain height to avoid pruned state crash

Both AH (56400→579775) and Relay (57100→708687) startBlocks were
pointing to blocks whose state has been pruned by the chain nodes.
This caused a crash loop: SubQuery couldn't index those blocks because
api.query calls returned 'State already discarded' errors.

Root cause: AH node has ~200 block state window (default pruning).
Fix: set startBlock to current tip - 50 blocks (safely within window).
Also added --state-pruning=archive to AH collator azad on VPS3 so
this doesn't recur as SubQuery re-syncs.

pezkuwi-assethub.yaml

@@ -19,7 +19,7 @@ network:
 dataSources:
   - name: main
     kind: substrate/Runtime
-    startBlock: 56400
+    startBlock: 579775
     mapping:
       file: ./dist/index.js
       handlers:

pezkuwi.yaml

@@ -22,7 +22,7 @@ network:
 dataSources:
   - name: main
     kind: substrate/Runtime
-    startBlock: 57100
+    startBlock: 708687
     mapping:
       file: ./dist/index.js
       handlers: