_gpu_dxg_adapterinfo_gate.nx

_gpu_dxg_adapterinfo_gate.nx -- SOVEREIGN-GPU ladder R1 (adapter identity), the HONEST re-scope. R0 proved raw reach (open /dev/dxg + LX_DXENUMADAPTERS2 -> the real GPU adapter set). R1 reads each adapter's REAL identity off the same raw fd via LX_DXQUERYADAPTERINFO (0xC0184709) type=15 ADAPTERTYPE -- the ONLY adapter-identity query the WSL2 dxgkrnl uapi exposes (there is NO description-STRING type on this kernel: kmtqueryadapterinfotype = {UMDRIVERPRIVATE=0, ADAPTERTYPE=15, ADAPTERTYPE_RENDER=57} only; the Windows adapter_string ABI is absent -- proven live by the R1 recon). POSITIVE ID (real, not a literal model string): the host returns a 4-byte adaptertype bitfield per adapter. The DISCRETE hardware GPU (hybrid_discrete=1, software_device=0) is the RTX 5080; the other is the integrated GPU (hybrid_integrated=1). This gate identifies the 5080 by that REAL bit, NOT by a hardcoded value. (Corrects the R0 caveat: both enumerated adapters are real paravirt hardware GPUs -- there is NO software-render adapter in this enum.) The human-readable "RTX 5080" model NAME is a LUID->PCI-vendor/device-id rung, deferred (dxgkrnl has no name string). NO-WAVE: zero throughput. GREEN iff (author=organ, from REAL device returns, never $?): A) enum ret==0 AND num_adapters>=2 AND every ADAPTERTYPE query ret==0; B) every adapter software_device==0 (all real hardware); C) EXACTLY ONE adapter is hybrid_discrete=1 (the 5080) and at least one is hybrid_integrated=1; D) the discrete value != the integrated value (not a stuck constant); E) TAMPER matrix all distinct from the real ret=0: T1 bogus type(8)!=0, T2 bogus ioctl code!=0, T3 query on a non-dxg fd(/dev/null)!=0. Marker -> knowledge/status/gpu_dxg.log (DXGINFOGATE). Raw syscalls only (no libvulkan/cuda/dxcore/libc). license_tier: ORIGINAL