Benchmark for F32 Sigmoid micro-kernels

- Add benchmark for F32 Sigmoid micro-kernels
- Rename Sigmoid micro-kernels to reflect ISA level and division algorithm

PiperOrigin-RevId: 281115949
diff --git a/src/init.c b/src/init.c
index f9922cf..5b87fe6 100644
--- a/src/init.c
+++ b/src/init.c
@@ -462,7 +462,7 @@
     };
     xnn_params.f32.clamp = (xnn_univector_ukernel_function) xnn_f32_clamp_ukernel__neon;
     xnn_params.f32.hswish = (xnn_univector_ukernel_function) xnn_f32_hswish_ukernel__neonfma;
-    xnn_params.f32.sigmoid = (xnn_univector_ukernel_function) xnn_f32_sigmoid_ukernel__neonfma_frac_p9_p10_x16;
+    xnn_params.f32.sigmoid = (xnn_univector_ukernel_function) xnn_f32_sigmoid_ukernel__neon_frac_p9_p10_nr1recps_x16;
     xnn_params.f32.prelu = (struct prelu_parameters) {
       .ukernel = (xnn_prelu_ukernel_function) xnn_f32_prelu_ukernel__neon_2x8,
       .row_tile = 2,