Reduce per-frame allocations: reuse emitter/order and reserve buffers in render state

[TiernanDeFranco]

Motivation

• Reduce frequent per-frame heap allocations and vector clears in particle preparation and render command draining to lower GC/allocator churn and improve runtime performance.

Description

• Add an emitter_order: Vec<usize> field to GpuPointParticles3D and reuse it across frames instead of allocating a temporary emitter_order vector each prepare call.
• Populate and sort self.emitter_order in prepare and iterate over it when pushing emitters, preserving capacity between frames.
• Stop clearing the render resource scratch vector in Runtime::drain_render_commands so the pre-allocated scratch capacity can be reused by resource_api::drain_commands.
• Add capacity reservations before Vec::append in RenderState::drain_commands and before appending pending transform roots in DirtyState::take_pending_transform_roots to avoid intermediate reallocations.

Testing

• Built the workspace with cargo build to verify no compilation errors, and the build succeeded.
• Ran the test suite with cargo test across affected crates to validate behavior, and the tests passed.

---

Codex Task