Lay groundwork for directly rendering to the canvas without a cpu roundrip (#1291)

* Add Texture handle type

* Add Texture View to shader inputs

* Implement basic rendering pipeline

* Render first texture using render pipeline

* Fix output color space

* Precompute the rendering pipeline

* Move gpu context creation to editor api

* Port gpu-executor nodes to node registry

* Fix canvas nodes and make code compile for non wasm targets

* Pin wasm-bindgen version

* Disable miri temoporarily for better ci times

* Fix formatting

* Remove unsafe block

* Bump wasm-pack version

* Bump wasm-bindgen version

* Add gpu feature guard for push node

* Make Into node async

node-graph/gcore/src/memo.rs

@@ -4,8 +4,8 @@ use core::future::Future;
 #[cfg(feature = "alloc")]
 use alloc::sync::Arc;
 use core::cell::Cell;
+use core::marker::PhantomData;
 use core::pin::Pin;
-use std::marker::PhantomData;
 
 // Caches the output of a given Node and acts as a proxy
 #[derive(Default)]
@@ -103,8 +103,6 @@ impl<'i, T: 'i + Clone> Node<'i, Option<T>> for LetNode<T> {
 	}
 }
 
-impl<T> std::marker::Unpin for LetNode<T> {}
-
 impl<T> LetNode<T> {
 	pub fn new() -> LetNode<T> {
 		LetNode { cache: Default::default() }