Finished scaffolding, lots of TODOs

- Most kernels just copy themselfs to get the shapes correct
- Matmul works only in 1 case and simply empty allocates otherwise
- Logits and randomized to make the demo finish itself.

Performance is quite bad (30ms/token), but lot's of prints and allocs and some actual sending to metal.

Couln't get it super high by removing the obvious blockers (println + the actual running matmuls).

Allocations takes between 1us and 100us and seems very stable, Maybe metal doesn't really have a smart allocator and we'll need to own it.

candle-nn/src/ops.rs

@@ -190,6 +190,16 @@ impl candle::CustomOp1 for SoftmaxLastDim {
             device: dev.clone(),
         };
         Ok((dst, layout.shape().clone()))
+    }    
+
+    #[cfg(feature = "metal")]
+    fn metal_fwd(
+        &self,
+        storage: &candle::MetalStorage,
+        layout: &Layout,
+    ) -> Result<(candle::MetalStorage, Shape)> {
+        println!("TODO softmax-last-dim");
+        Ok((storage.clone(), layout.shape().clone()))
     }
 }