Avoid using flume::Receiver::into_stream() to avoid memory leaks until the issue is resolved upstream (#394)

* Tweak rolling_total test to also confirm capacity doesn't go nuts

* Use Jemalloc

* Avoid flume's into_stream and use a workaround for now

* cargo fmt

* Improve comments now that there's an issue to point to

backend/telemetry_shard/Cargo.toml

@@ -25,3 +25,6 @@ structopt = "0.3.21"
 thiserror = "1.0.25"
 tokio = { version = "1.10.1", features = ["full"] }
 tokio-util = { version = "0.6", features = ["compat"] }
+
+[target.'cfg(not(target_env = "msvc"))'.dependencies]
+jemallocator = "0.3.2"

backend/telemetry_shard/src/main.rs

@@ -35,6 +35,13 @@ use hyper::{Method, Response};
 use simple_logger::SimpleLogger;
 use structopt::StructOpt;
 
+#[cfg(not(target_env = "msvc"))]
+use jemallocator::Jemalloc;
+
+#[cfg(not(target_env = "msvc"))]
+#[global_allocator]
+static GLOBAL: Jemalloc = Jemalloc;
+
 const VERSION: &str = env!("CARGO_PKG_VERSION");
 const AUTHORS: &str = env!("CARGO_PKG_AUTHORS");
 const NAME: &str = "Substrate Telemetry Backend Shard";