Add indirect compute dispatch support for WebGPU

[mvaligursky]

Fixes #8000

Add indirect compute dispatch support for WebGPU

Adds support for indirect compute dispatch on WebGPU, allowing compute shaders to generate dispatch parameters for subsequent compute shaders. This enables GPU-driven workloads where the number of workgroups is determined dynamically on the GPU without CPU readback.

New Public API

This closely resembles existing indirect drawing API.

GraphicsDevice

Property/Method	Description
maxIndirectDispatchCount	Maximum number of indirect dispatch slots per frame (default: 256)
getIndirectDispatchSlot(count?)	Reserve slot(s) in the indirect dispatch buffer for the current frame
indirectDispatchBuffer	Storage buffer for indirect dispatch parameters (WebGPU only, null on other platforms)

Compute

Method	Description
setupIndirectDispatch(slotIndex, buffer?)	Configure compute to read dispatch parameters from a buffer

Usage

// Reserve a slot in the indirect dispatch buffer
const slot = device.getIndirectDispatchSlot();

// First compute shader writes dispatch parameters to the buffer
prepareCompute.setParameter('indirectBuffer', device.indirectDispatchBuffer);
prepareCompute.setParameter('slot', slot);
prepareCompute.setupDispatch(1, 1, 1);
device.computeDispatch([prepareCompute]);

// Second compute shader uses indirect dispatch
processCompute.setupIndirectDispatch(slot);
device.computeDispatch([processCompute]);## Technical Details

• Each indirect dispatch slot contains 3 × 32-bit values (x, y, z workgroup counts)
• Built-in buffer slots are frame-stamped and validated to ensure setupIndirectDispatch is called each frame
• Custom storage buffers can be provided for complex scheduling outside rendering frames

Example

New indirect-dispatch example demonstrates depth-based edge detection:

1. Scan shader analyzes depth buffer and classifies tiles by depth discontinuity
2. Writes tile counts to indirect dispatch buffer
3. Effect shaders are indirectly dispatched to process only the relevant tiles