ggml-webgpu: add q4_0/q8_0 SET_ROWS (#23760)

* Add q8_0 and q4_0 set_rows

* Add fast(er) quantization set_rows path

* formatting/naming

* a little more naming

* Remove unused constant

* Don't override other override

* Avoid bitcast

* Narrow relaxation

tests/test-backend-ops.cpp

@@ -2415,6 +2415,15 @@ struct test_set_rows : public test_case {
         }
         return 1e-7;
     }
+
+    // See dicussion here: https://github.com/ggml-org/llama.cpp/pull/23760#issuecomment-4566312209
+    double max_nmse_err(ggml_backend_t backend) override {
+        ggml_backend_reg_t reg = ggml_backend_dev_backend_reg(ggml_backend_get_device(backend));
+        if (type == GGML_TYPE_Q8_0 && strcmp(ggml_backend_reg_name(reg), "WebGPU") == 0) {
+            return std::max(test_case::max_nmse_err(backend), 2e-7);
+        }
+        return test_case::max_nmse_err(backend);
+    }
 };
 
 // GGML_OP_ROPE + GGML_OP_VIEW + GGML_OP_SET_ROWS