Reduce `IntPtr` Overhead in Native Interface Wrapper

[patdhlk]

Summary

The current C# wrapper around the native interface relies heavily on IntPtr usage, which introduces unnecessary overhead in performance‑critical paths. This affects both call latency and memory churn, especially when repeatedly converting pointers, allocating unmanaged memory, or performing redundant marshaling.

Problem Details

• Frequent IntPtr → delegate conversions using Marshal.GetDelegateForFunctionPointer
• Repeated calls into native code instead of batching operations
• Use of non‑blittable structs causing marshaling overhead
• Unnecessary allocations via Marshal.AllocHGlobal and Marshal.Copy
• Delegates used where function pointers (delegate* unmanaged) would be faster
• Potential boxing of IntPtr values in wrapper classes or collections
• SafeHandle usage in hot paths adding reference counting and finalization overhead

Proposed Improvements

• Cache all function pointers at initialization time
• Replace delegates with delegate* unmanaged where possible
• Ensure structs passed to native code are blittable
• Pin managed buffers instead of allocating unmanaged memory
• Reduce managed↔native transitions by batching operations
• Avoid SafeHandle in performance‑critical paths when lifetime is already controlled
• Use [NativeLibrary.Load](https://learn.microsoft.com/en-us/dotnet/api/system.runtime.interopservices.nativelibrary.load?view=net-9.0) and NativeLibrary.GetExport for explicit control over symbol resolution

Expected Benefits

• Lower call overhead for native function invocations
• Reduced GC pressure from fewer allocations
• Faster throughput in tight loops or high‑frequency native calls
• More predictable performance characteristics