Add manually-runnable benchmarks for runtime profiling (#2005)

* Split benches into two files

* Implement executor update bench

* Restructure benchmarks

* Unify usages of wrap network in scope

* Remove unused imports

* Fix oom bug

* Remove bounding box impl

node-graph/graph-craft/benches/compile_demo_art_criterion.rs

@@ -0,0 +1,14 @@
+use criterion::{black_box, criterion_group, criterion_main, Criterion};
+use graph_craft::util::DEMO_ART;
+fn compile_to_proto(c: &mut Criterion) {
+	use graph_craft::util::{compile, load_from_name};
+	let mut c = c.benchmark_group("Compile Network cold");
+
+	for name in DEMO_ART {
+		let network = load_from_name(name);
+		c.bench_function(name, |b| b.iter_batched(|| network.clone(), |network| compile(black_box(network)), criterion::BatchSize::SmallInput));
+	}
+}
+
+criterion_group!(benches, compile_to_proto);
+criterion_main!(benches);