Initial extraction from isis: direct /dev/kfd GPU compute

AMD RDNA3 GPU driver via raw KFD ioctls. No ROCm, no OpenCL.

- Device discovery and topology queries
- VRAM/GTT/Userptr memory with shared address space
- PM4 compute queue and kernel dispatch
- Hand-written RDNA3 ASM kernels: matmul (3100 GFLOP/s), matvec, superlinear
- Tile-safe buffer padding for OOB protection
- Event-based interrupt wait (no CPU polling)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Cargo.toml

@@ -0,0 +1,12 @@
+[package]
+name = "kfd"
+version = "0.1.0"
+edition = "2021"
+description = "Direct /dev/kfd GPU compute for AMD RDNA3 — no ROCm, no OpenCL, just ioctl"
+license = "MIT OR Apache-2.0"
+repository = "https://git.rotko.net/tommi/kfd"
+
+[dependencies]
+libc = "0.2"
+
+[dev-dependencies]