Reduce proximity cache startup grace period from 30s to 5s

The 30-second startup grace period was causing test flakiness in
`test_multiple_clients_subscription`. The test would fail because:

1. Node A caches a contract during startup but delays the proximity
   cache announcement for 30 seconds
2. Node B tries to GET the contract before the announcement is sent
3. Without the proximity cache information, Node B can't locate the
   contract and times out

Reducing the grace period to 5 seconds:
- Still provides batching benefits during rapid contract loading
- Prevents long delays in cross-node GET operations
- Fixes the flaky test that was timing out after 8+ minutes in CI

The 5-second window is sufficient for batching announcements during
the rapid initialization phase while not causing excessive delays
for critical operations.

Fixes test_multiple_clients_subscription timeout issue.

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

Co-Authored-By: Claude <[email protected]>

Author: sanity

crates/core/src/node/proximity_cache.rs

@@ -92,7 +92,7 @@ impl ProximityCacheManager {
         &self,
         contract_key: &ContractKey,
     ) -> Option<ProximityCacheMessage> {
-        const STARTUP_GRACE_PERIOD: Duration = Duration::from_secs(30);
+        const STARTUP_GRACE_PERIOD: Duration = Duration::from_secs(5);
 
         let contract_id = *contract_key.id();
 
@@ -202,7 +202,7 @@ impl ProximityCacheManager {
         use crate::config::GlobalExecutor;
 
         GlobalExecutor::spawn(async move {
-            const STARTUP_GRACE_PERIOD: Duration = Duration::from_secs(30);
+            const STARTUP_GRACE_PERIOD: Duration = Duration::from_secs(5);
 
             // Wait for startup grace period to complete
             tokio::time::sleep(STARTUP_GRACE_PERIOD).await;