Improve volume sampling and fix rendering issues

[tarcila]

• Nearest neighbour interpolation for NanoVDB volumes
• Volume gradient/normal computation across all spatial field types for shading support
• MDL fix for texture resource null-pointer dereference and uninitialized memory access
• Infinite accumulation when unitDistance is 0 in transfer function volumes
• Unused maxRayDepth from default/directLight renderer JSON

[Copilot]

Pull request overview

Improves volume rendering correctness and shading by adding gradient/normal support to spatial field sampling, introducing NanoVDB nearest-neighbor filtering, and fixing a few robustness issues (MDL sampler lifetime, unitDistance edge cases, renderer JSON cleanup).

Changes:

• Add optional gradient output to volume samplers and use it to compute world-space volume normals for the normal AOV.
• Add NanoVDB nearest/linear filtering selection and plumb it through GPU data + PTX samplers.
• Fix null/uninitialized MDL sampler handling; avoid division-by-zero/infinite accumulation for unitDistance.

Reviewed changes

Copilot reviewed 17 out of 17 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
devices/rtx/libmdl/types.h	Initialize TextureDescriptor::knownIndex to avoid uninitialized reads.
devices/rtx/device/volume/TransferFunction1D.cpp	Clamp oneOverUnitDistance defensively; add warning for invalid unitDistance.
devices/rtx/device/visrtx_device.json	Remove unused maxRayDepth renderer parameter from schema.
devices/rtx/device/spatial_field/StructuredRegularSampler_ptx.cu	Add optional gradient computation via central differences.
devices/rtx/device/spatial_field/StructuredRectilinearSampler_ptx.cu	Refactor sampling helper and add gradient computation.
devices/rtx/device/spatial_field/NvdbRegularSampler_ptx.cu	Add nearest/linear sampling and optional gradients; adjust index mapping.
devices/rtx/device/spatial_field/NvdbRegularField.cpp	Plumb NanoVDB filter choice into GPU data.
devices/rtx/device/spatial_field/NvdbRectilinearSampler_ptx.cu	Add nearest/linear sampling and optional gradients.
devices/rtx/device/spatial_field/NvdbRectilinearField.cpp	Plumb NanoVDB filter choice into GPU data.
devices/rtx/device/material/MDL.h	Default-initialize sampler descriptor members to safe values.
devices/rtx/device/material/MDL.cpp	Guard against null samplers and warn on failed default texture acquisition.
devices/rtx/device/gpu/volumeIntegration.h	Add sample-with-gradient path; compute volume normals at first contribution depth.
devices/rtx/device/gpu/shadingState.h	Extend NanoVDB sampler state to support nearest/linear samplers + filter flag.
devices/rtx/device/gpu/renderer/raygen_helpers.h	Accumulate volume normals from gradients; normalize material normals before accumulation.
devices/rtx/device/gpu/populateHit.h	Populate VolumeHit transform matrices from OptiX instance handle.
devices/rtx/device/gpu/gpu_objects.h	Introduce SpatialFieldFilter and store it in NanoVDB GPU data.
devices/rtx/device/gpu/gpu_math.h	Add worldToObject/objectToWorld matrices to VolumeHit.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[jeffamstutz]

LGTM, thanks!