Handle errors gathering metrics from slaves

If the session to a slave is lost while gathering metrics the error
would propagate out to the metrics route handler and break collection of
statistics.  Fix by handling errors thrown from sending the getMetrics
request to slaves.

packages/master/src/routes.js

@@ -8,6 +8,7 @@ const jwt = require("jsonwebtoken");
 const path = require("path");
 const util = require("util");
 
+const libErrors = require("@clusterio/lib/errors");
 const libHelpers = require("@clusterio/lib/helpers");
 const libLink = require("@clusterio/lib/link");
 const libPlugin = require("@clusterio/lib/plugin");
@@ -47,7 +48,15 @@ async function getMetrics(req, res, next) {
 	let requests = [];
 	let timeout = req.app.locals.master.config.get("master.metrics_timeout") * 1000;
 	for (let slaveConnection of req.app.locals.master.wsServer.slaveConnections.values()) {
-		requests.push(libHelpers.timeout(libLink.messages.getMetrics.send(slaveConnection), timeout, null));
+		requests.push(libHelpers.timeout(
+			libLink.messages.getMetrics.send(slaveConnection).catch(err => {
+				if (!(err instanceof libErrors.SessionLost)) {
+					logger.error(`Unexpected error gathering metrics from slave:\n${err.stack}`);
+				}
+				return null;
+			}),
+			timeout, null
+		));
 	}
 
 	for await (let result of await libPrometheus.defaultRegistry.collect()) {