Move entity model streaming and polling to worker threads

[iManGaaX]

Summary

• Refactored CClientModelRequestManager to process model streaming requests asynchronously via worker threads.
• Implemented a dual-mutex architecture utilizing m_Mutex for container safety and m_ModelInfoMutex to serialize game engine and RenderWare calls.
• Added a thread-safe cancellation mechanism (bCancelled) ensuring all memory cleanup and engine callbacks remain strictly on the main thread inside DoPulse()

Motivation

• Eliminates severe frame drops and game stuttering (FPS stutters) caused by synchronous disk I/O and heavy model loading operations on the main thread.

Test plan

1. Compiled the MTA client source code locally.
2. Conducted stress tests by rapidly spawning and destroying custom entities and models to simulate heavy streaming loads.
3. Verified stable frame rates, smooth movement, and complete absence of use-after-free crashes or deadlocks during dynamic request cancellations.

Checklist

• Your code should follow the coding guidelines.
• Smaller pull requests are easier to review. If your pull request is beefy, your pull request should be reviewable commit-by-commit.