feat(quantized): VNNI INT8 GEMM via VPDPBUSD (#128, sprint W3-C)

Closes parity item 12 — INT8 GEMM accelerated via AVX-512 VNNI's VPDPBUSD
instruction (4-element u8×i8→i32 dot product). Falls back to scalar
int8_gemm_i32 on hardware without VNNI.

What ships:
- src/hpc/vnni_gemm.rs (387 LOC): int8_gemm_vnni public API,
  has_vnni() detection, _mm512_dpbusd_epi32 inner kernel, scalar fallback
- src/hpc/simd_caps.rs: avx512vnni: bool field added to SimdCaps,
  is_x86_feature_detected!("avx512vnni") detection wired
- src/hpc/mod.rs: pub mod vnni_gemm declaration

Hardware coverage:
- AVX-512 VNNI: Ice Lake, Sapphire Rapids, Zen 4 (with AVX-512), Tiger Lake
- Fallback: any x86_64 / ARM / scalar

Tests: 11 passing (4×4, 16×16, 17×17 tail, 1×1 edge, mixed values).
Total lib tests: 1817+ pass.

Note: type-cast fix applied to _mm512_loadu_si512 / _mm512_storeu_si512
(*const i32 → *const __m512i, *mut i32 → *mut __m512i) per Rust 1.94
intrinsic signatures.

https://claude.ai/code/session_01NYGrxVopyszZYgLBxe4hgj

Co-authored-by: Claude <noreply@anthropic.com>

Author: AdaWorldAPI

src/hpc/mod.rs

@@ -411,3 +411,4 @@ mod e2e_tests {
         assert!(bnn_result.score > -1.0 && bnn_result.score < 1.0);
     }
 }
+pub mod vnni_gemm;

src/hpc/simd_caps.rs

@@ -41,6 +41,9 @@ pub struct SimdCaps {
     pub sse2: bool,
     /// FMA (fused multiply-add).
     pub fma: bool,
+    /// AVX-512 VNNI (VPDPBUSD — u8×i8→i32 dot product of 4-element groups).
+    /// Present on Ice Lake, Sapphire Rapids, Zen 4 (with AVX-512), Tiger Lake.
+    pub avx512vnni: bool,
 
     // ── aarch64 (ARM) ──
     /// NEON 128-bit SIMD (mandatory on aarch64, always true).
@@ -82,6 +85,7 @@ impl SimdCaps {
             sse41: is_x86_feature_detected!("sse4.1"),
             sse2: is_x86_feature_detected!("sse2"),
             fma: is_x86_feature_detected!("fma"),
+            avx512vnni: is_x86_feature_detected!("avx512vnni"),
             // ARM fields: all false on x86
             neon: false,
             asimd_dotprod: false,
@@ -107,6 +111,7 @@ impl SimdCaps {
             sse41: false,
             sse2: false,
             fma: false,
+            avx512vnni: false,
             // ARM fields: runtime detection
             neon: true, // mandatory on aarch64
             asimd_dotprod: std::arch::is_aarch64_feature_detected!("dotprod"),
@@ -129,6 +134,7 @@ impl SimdCaps {
             sse41: false,
             sse2: false,
             fma: false,
+            avx512vnni: false,
             neon: false,
             asimd_dotprod: false,
             fp16: false,
@@ -150,6 +156,13 @@ impl SimdCaps {
         self.avx512bw && self.avx512vpopcntdq
     }
 
+    /// True if AVX-512 VNNI is available (VPDPBUSD on zmm registers).
+    /// Present on Ice Lake, Tiger Lake, Sapphire Rapids, Zen 4.
+    #[inline(always)]
+    pub fn has_avx512_vnni(self) -> bool {
+        self.avx512f && self.avx512vnni
+    }
+
     // ── ARM convenience methods ──
 
     /// True if running on aarch64 with NEON (always true on aarch64).