Cover PSO-only RT surface: recursion, RAY_FLAG_SKIP_CLOSEST_HIT_SHADER, callables

[MarijnS95]

Three tests stacked on #1275 covering features that inline RT (RayQuery in compute) physically can't express — they're only reachable through a DispatchRays-driven RT pipeline.

• trace-ray-recursion.test — closest-hit fires a secondary TraceRay from an above-the-triangle origin. First-level CH sees payload=0 → bumps to 0x1 → calls TraceRay. Second-level CH sees payload!=0 → writes 0x10. Unwinds: first-level OR's in 0x100. Final payload 0x110 (272 decimal). RayTracingPipelineConfig.MaxTraceRecursionDepth: 2 so both TraceRay calls are within budget.

• ray-flag-skip-closest-hit.test — two lanes fire identical rays at the same triangle. Lane 0 uses RAY_FLAG_NONE so CH runs and writes 0xBEEF. Lane 1 uses RAY_FLAG_SKIP_CLOSEST_HIT_SHADER (PSO-only — inline RT has no equivalent) so CH is skipped and payload keeps its initial 0xAAAA. Output [0xBEEF, 0xAAAA].

• callable-shader.test — two callable shaders writing distinct sentinels (0xAAAA / 0xBBBB). Each lane calls CallShader(Idx, ...) so the SBT callable region's per-record routing is exercised independently of the hit-group / miss routing already covered in Cover DispatchRays index/dimensions + SBT miss/hit-group routing #1277. Callable shaders themselves don't exist in inline RT.

  Only # XFAIL: Clang — the clang HLSL path doesn't implement the RT pipeline yet ([EPIC]: Bring-up and test coverage for PSO-based raytracing #1268). On Vulkan the current DXC -spirv output is valid: each callable's IncomingCallableDataKHR is listed only in its own entry point's interface and spirv-val is clean, so the module loads and the test passes (confirmed by an XPASS on the windows-intel VK CI runner). An earlier DXC over-listed those interface variables in every callable entry point, which has since been fixed upstream — so no Vulkan XFAIL or spirv-opt cleanup is needed anymore.

All three pass on Metal once the Metal bring-up (#1281) sets setMaxCallStackDepth(MaxTraceRecursionDepth) so nested TraceRay actually unwinds; with the default of 1 the second TraceRay was silently dropped and the recursion test produced 0x1 instead of 0x110. With #1281 rebased underneath, Metal is dropped from the XFAIL list on all three tests.

Test plan

Local on an NVIDIA RTX 3060:

• Linux Vulkan (native offloader)
• Linux D3D12 (Wine + vkd3d-proton + cross-compiled offloader.exe)
• Windows Vulkan (native offloader.exe)
• Windows D3D12 (native offloader.exe)

CI (RT-capable runners):

• windows-nvidia D3D12 (RaytracingTier 1.2)
• windows-intel VK (VK_KHR_ray_tracing_pipeline) — all three RT tests pass; callable previously surfaced as an XPASS against the stale Vulkan XFAIL that this revision removes
• macOS Metal (supportsRaytracing)

[MarijnS95]

Only lavapipe fails specifically on all RT tests and existing InlineRT tests.