Fix realtime message and call latency

[smshagor-dev]

Summary

Removes the main first-send, reconnect, and first-call latency sources while preserving SyncChat's existing reliability, security, moderation, E2EE, and durable call-state behavior.

Messaging

• Connect the authenticated Socket.IO client immediately and use faster reconnect backoff.
• Reuse the already-warm Socket.IO Redis connection for chat abuse/rate-limit checks instead of opening a new Redis connection on the first message.
• Prewarm realtime backend dependencies at startup.
• Run safe chat sequence/mention work concurrently and persist reliability metadata in the original chat create instead of a second write.
• Ack durable messages earlier, before non-critical mention/message-request side effects.
• Add optimistic sender-side message rendering with clientMessageId reconciliation, failure state, and durable outbox compatibility.

Calling

• Reuse the warm Redis runtime for durable call state instead of connecting Redis lazily on the first call.
• Prewarm calling configuration on both backend and client.
• Coalesce concurrent calling-config requests.
• Start validated server-side call reservation/ringing as soon as the outgoing call UI opens, so camera/microphone acquisition no longer blocks the recipient notification path.
• Add duplicate-safe call/start coordination and initiator catch-up so a fast recipient cannot race ahead of caller media initialization.

Validation scope

• Backend build explicitly syntax-checks the new call realtime coordinator.
• No API contracts, room permissions, call policy, message moderation, E2EE, or durable outbox guarantees are removed.

[vercel]

Deployment failed for project syncchat with the following error:

Resource is limited - try again in 24 hours (more than 100, code: "api-deployments-free-per-day").

Learn More: https://vercel.com/smshagor-devs-projects?upgradeToPro=build-rate-limit