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/interpreted-executor/benches/run_once.rs

@@ -0,0 +1,24 @@
+use criterion::{criterion_group, criterion_main, Criterion};
+use graph_craft::graphene_compiler::Executor;
+use graphene_std::transform::Footprint;
+
+mod benchmark_util;
+use benchmark_util::{bench_for_each_demo, setup_network};
+
+fn run_once(c: &mut Criterion) {
+	let mut group = c.benchmark_group("Run Once");
+	let footprint = Footprint::default();
+	bench_for_each_demo(&mut group, |name, g| {
+		g.bench_function(name, |b| {
+			b.iter_batched(
+				|| setup_network(name),
+				|(executor, _)| futures::executor::block_on((&executor).execute(criterion::black_box(footprint))),
+				criterion::BatchSize::SmallInput,
+			)
+		});
+	});
+	group.finish();
+}
+
+criterion_group!(benches, run_once);
+criterion_main!(benches);