[Feature] Compute pipeline creation

[vmarcella]

Overview

Add support for compute pipelines to enable GPU compute operations for
particle simulation, culling, and general-purpose computation.

Current State

No response

Scope

Goals:

• Compute shader loading and compilation
• Compute pipeline builder
• Workgroup size configuration
• Pipeline caching

Non-Goals:

• Storage buffers
• Dispatch commands
• Async compute queues

Proposed API

// crates/lambda-rs/src/render/compute.rs
pub struct ComputePipeline {
  pipeline: wgpu::ComputePipeline,
  workgroup_size: [u32; 3],
}

pub struct ComputePipelineBuilder {
  shader: Option<ShaderModule>,
  entry_point: String,
  bind_group_layouts: Vec<BindGroupLayout>,
}

impl ComputePipelineBuilder {
  pub fn new() -> Self;
  pub fn with_shader(self, shader: &ShaderModule) -> Self;
  pub fn with_entry_point(self, name: &str) -> Self;
  pub fn with_bind_group_layout(self, layout: &BindGroupLayout) -> Self;
  pub fn build(self, gpu: &Gpu) -> Result<ComputePipeline, Error>;
}

impl ComputePipeline {
  pub fn workgroup_size(&self) -> [u32; 3];
}

Acceptance Criteria

• Compute shaders compile successfully
• Pipeline creation with custom bind groups
• Workgroup size accessible
• Error handling for invalid shaders
• Example with simple compute shader

Affected Crates

lambda-rs, lambda-rs-platform

Notes

• Entry point typically "main" for compute
• Workgroup size declared in shader, queried from reflection