perf: use fast_activation instead of broadcasting

Author: avik-pal

bench/comparison.md

@@ -6,11 +6,11 @@
 
 | #layers | Forward         | Train: 10 epochs     |
 | ------- | --------------- | -------------------- |
-| 1       | 21.9836 ms      | 785.9677 ms          |
-| 2       | 35.2872 ms      | 1340.3389 ms         |
-| 3       | 51.0229 ms      | 2105.0524 ms         |
-| 4       | 54.9812 ms      | 2746.7931 ms         |
-| 5       | 64.6753 ms      | 3043.4466 ms         |
+| 1       | 13.215269 ms    | 857.77145 ms         |
+| 2       | 21.864486 ms    | 1509.3964 ms         |
+| 3       | 29.911771 ms    | 2039.654284 ms       |
+| 4       | 37.942865 ms    | 2417.136518 ms       |
+| 5       | 43.498116 ms    | 2822.866553 ms       |
 
 ### Flux.jl (Julia)