Revert the runner-side idle timeout to 1s (#1483)

We stumbled over a problem where *some* FDKs are idling their UDS HTTP
connections at periods lower than the 120s that this was expecting.

This was giving rise to spurious errors (from older versions of the node
FDK, for instance) - where invocations beyond the first were seeing
502 gateway errors from a prematurely-closed UDS socket.

The notion of an idle timeout here is good, but we should check that all
FDKs have the appropriate behaviour and give users time to rev their
functions before reintroducing this.

Author: jan-g

api/agent/agent.go

@@ -1207,7 +1207,8 @@ func newHotContainer(ctx context.Context, evictor Evictor, call *call, cfg *Conf
 		MaxIdleConns:           1,
 		MaxIdleConnsPerHost:    1,
 		MaxResponseHeaderBytes: int64(cfg.MaxHdrResponseSize),
-		IdleConnTimeout:        120 * time.Second, // TODO(reed): since we only allow one, and we close them, this is gratuitous?
+		IdleConnTimeout:        1 * time.Second, // TODO(jang): revert this to 120s at the point all FDKs are known to be fixed
+		// TODO(reed): since we only allow one, and we close them, this is gratuitous?
 		DialContext: func(ctx context.Context, _, _ string) (net.Conn, error) {
 			var d net.Dialer
 			return d.DialContext(ctx, "unix", filepath.Join(iofs.AgentPath(), udsFilename))