Add avg_pool2d metal implementation for the metal backend (#1869)

* implement metal avg pool 2d

* fixX

* add suggested precision workaround for the accumulator

candle-metal-kernels/src/lib.rs

@@ -1827,7 +1827,7 @@ fn divide(m: usize, b: usize) -> NSUInteger {
 }
 
 #[allow(clippy::too_many_arguments)]
-pub fn call_max_pool2d(
+pub fn call_pool2d(
     device: &Device,
     command_buffer: &CommandBufferRef,
     kernels: &Kernels,