gl: Improve CS throughput

- Avoids making too many invocations, especially given the 1D nature of some GPU dispatch handlers

rpcs3/Emu/RSX/GL/GLCompute.cpp

@@ -21,6 +21,8 @@ namespace gl
 			optimal_group_size = 128;
 		}
 
+		optimal_kernel_size = 256 / optimal_group_size;
+
 		glGetIntegeri_v(GL_MAX_COMPUTE_WORK_GROUP_COUNT, 0, reinterpret_cast<GLint*>(&max_invocations_x));
 	}