Optimization wireframe index generation

[AlexAPPi]

Description

Replaced dynamic buffer allocation with pre-typed memory structures
to minimize runtime overhead and reduce GC pressure. This improves
wireframe index generation performance by eliminating type checks
and avoiding frequent temporary allocations.

I have read the contributing guidelines

[Copilot]

Pull Request Overview

This PR optimizes wireframe index generation by replacing dynamic array allocation with pre-typed memory structures to reduce runtime overhead and GC pressure.

Key changes:

• Replaced dynamic lines array with pre-allocated typed arrays
• Eliminated intermediate array operations by writing directly to typed array buffers
• Optimized buffer creation by passing the typed array buffer directly to IndexBuffer constructor

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[mvaligursky]

AI review, which it seems to make sense after a quick look:

Assessment
Pros:
✅ Eliminates dynamic array growth overhead
✅ Removes one copy operation (direct buffer passing)
✅ Better for performance-critical scenarios
✅ Handles edge case where vertices aren't divisible by 3
Concerns:
⚠️ Indexed path memory inefficiency: srcIndices.slice(0) allocates a full copy of the source index buffer even though typically only a fraction is needed (unique edges << total indices)
⚠️ Code is less readable
⚠️ For small meshes, the optimization benefit may be negligible compared to complexity cost
Recommendation
The non-indexed path optimization is excellent and should definitely be applied.
For the indexed path, I'd suggest an alternative: pre-allocate based on worst-case (count * 2) instead of copying the entire source buffer:
This would be more memory-efficient while keeping the performance benefits.