i32 mat4 multiply-add

x = m * x + c chained in a register, where m is a 4x4 integer matrix emulated as four vec4<i32> rows combined with dot() (WGSL has no mat4x4<i32>): the same 4 dot products a real int4x4 multiply compiles to.

Measured in OP/s. Higher values mean better performance. Benchmark your device.