Revert: re-enable MCP servers on staging and production (#5022)

## Summary
- Reverts #5020 which set `replicaCount: 0` for MCP servers
- Restores `replicaCount: 2` in both production and staging values

## Test plan
- [ ] Verify MCP pods come back after merge: `kubectl get pods -n
everyrow-mcp` and `kubectl get pods -n everyrow-mcp-staging`

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Sourced from commit 124d2db62ef28ee71991f2bffd9e52e28828ae0b

Author: RafaelPo

futuresearch-mcp/deploy/chart/values.staging.yaml

@@ -6,7 +6,7 @@
 # FUTURESEARCH_API_URL is intentionally not overridden — there is no staging API.
 # The MCP server is a frontend to the production FutureSearch API in both environments.
 
-replicaCount: 0
+replicaCount: 2
 
 env:
   MCP_SERVER_URL: "https://mcp-staging.futuresearch.ai"

futuresearch-mcp/deploy/chart/values.yaml

@@ -9,7 +9,7 @@ image:
 releaseId: "latest"
 
 # OAuth state is stored in Redis. Multiple replicas are possible.
-replicaCount: 0
+replicaCount: 2
 
 service:
   port: 80