Restructure GPU execution to model GPU pipelines in the node graph (#1088)

* Start implementing GpuExecutor for wgpu

* Implement read_output_buffer function

* Implement extraction node in the compiler

* Generate type annotations during shader compilation

* Start adding node wrapprs for graph execution api

* Wrap more of the api in nodes

* Restructure Pipeline to accept arbitrary shader inputs

* Adapt nodes to new trait definitions

* Start implementing gpu-compiler trait

* Adapt shader generation

* Hardstuck on pointer casts

* Pass nodes as references in gpu code to avoid zsts

* Update gcore to compile on the gpu

* Fix color doc tests

* Impl Node for node refs

node-graph/gcore/src/storage.rs

@@ -0,0 +1,34 @@
+use crate::Node;
+
+use core::marker::PhantomData;
+use core::ops::{DerefMut, Index, IndexMut};
+
+struct SetNode<S, I, Storage, Index> {
+	storage: Storage,
+	index: Index,
+	_s: PhantomData<S>,
+	_i: PhantomData<I>,
+}
+
+#[node_macro::node_fn(SetNode<_S, _I>)]
+fn set_node<T, _S, _I>(value: T, storage: &'any_input mut _S, index: _I)
+where
+	_S: IndexMut<_I>,
+	_S::Output: DerefMut<Target = T> + Sized,
+{
+	*storage.index_mut(index).deref_mut() = value;
+}
+
+struct GetNode<S, Storage> {
+	storage: Storage,
+	_s: PhantomData<S>,
+}
+
+#[node_macro::node_fn(GetNode<_S>)]
+fn get_node<_S, I>(index: I, storage: &'any_input _S) -> &'input _S::Output
+where
+	_S: Index<I>,
+	_S::Output: Sized,
+{
+	storage.index(index)
+}