ggml-webgpu: updated matrix-vector multiplication (#21738)

* merged properly, but slow q3_k and q5_k with u32 indexing

* Start on new mat-vec

* New format float paths working

* Working q4_0

* Work on remaining legacy q-types

* port k-quants to new matvec

* remove old shader

* Remove old constants, format

* remove accidental file

---------

Co-authored-by: Neha Abbas <nehaabbas@ReeseLevines-MacBook-Pro.local>
Co-authored-by: Reese Levine <reeselevine1@gmail.com>

ggml/src/ggml-webgpu/wgsl-shaders/common_decls.tmpl

@@ -45,6 +45,13 @@ fn load_u16_at_src0(byte_offset: u32) -> u32 {
     return (word >> shift) & 0xFFFFu;
 }
 
+// Always reads the 4-byte-aligned word containing byte_offset.
+// Caller extracts the 16-bit half it needs via & 0xFFFFu or >> 16u.
+// this is used in k-quants for better performance
+fn load_u32_at_src0_aligned(byte_offset: u32) -> u32 {
+    return src0[(byte_offset & ~3u) / 4u];
+}
+
 fn load_u32_at_src0(byte_offset: u32) -> u32 {
     let word_idx = byte_offset / 4u;
     let shift = (byte_offset & 0x3u) * 8u;