Refactor the kraken module to automatically re-connect on errors

In order to be able to re-connect on certain errors, we model
connection errors separately from parsing errors. We also change
the API of the whole module to no longer forward all errors to
the subscribers but instead, only update the subscribers with
either a latest rate or a permanent failure in case we exhausted
all our options to re-connect the websocket.

To model all of this properly, we introduce to sub-modules so that
each submodule can have their own `Error` type.

Resolves #297.

swap/src/bin/asb.rs

@@ -92,7 +92,7 @@ async fn main() -> Result<()> {
                 bitcoin_wallet.new_address().await?
             );
 
-            let kraken_rate_updates = kraken::connect().await?;
+            let kraken_rate_updates = kraken::connect()?;
 
             let (event_loop, _) = EventLoop::new(
                 config.network.listen,

swap/src/bin/kraken_ticker.rs

@@ -3,12 +3,10 @@ use anyhow::{Context, Result};
 #[tokio::main]
 async fn main() -> Result<()> {
     tracing::subscriber::set_global_default(
-        tracing_subscriber::fmt().with_env_filter("trace").finish(),
+        tracing_subscriber::fmt().with_env_filter("debug").finish(),
     )?;
 
-    let mut ticker = swap::kraken::connect()
-        .await
-        .context("Failed to connect to kraken")?;
+    let mut ticker = swap::kraken::connect().context("Failed to connect to kraken")?;
 
     loop {
         match ticker.wait_for_update().await? {