Add a commands-based interface to Websocket handlers

This feature is currently experimental. It will become the
preferred way to use Websocket handlers once it becomes
documented.

A commands-based interface enables adding commands without
having to change the interface much. It mirrors the interface
of stream handlers or gen_statem. It will enable adding
commands that have been needed for some time but were not
implemented for fear of making the interface too complex.

test/handlers/ws_init_commands_h.erl

@@ -0,0 +1,30 @@
+%% This module takes commands from the x-commands header
+%% and returns them in the websocket_init/1 callback.
+
+-module(ws_init_commands_h).
+-behavior(cowboy_websocket).
+
+-export([init/2]).
+-export([websocket_init/1]).
+-export([websocket_handle/2]).
+-export([websocket_info/2]).
+
+init(Req=#{pid := Pid}, RunOrHibernate) ->
+	Commands0 = cowboy_req:header(<<"x-commands">>, Req),
+	Commands = binary_to_term(base64:decode(Commands0)),
+	case Commands of
+		bad -> ct_helper_error_h:ignore(Pid, cowboy_websocket, handler_call, 6);
+		_ -> ok
+	end,
+	{cowboy_websocket, Req, {Commands, RunOrHibernate}}.
+
+websocket_init(State={Commands, run}) ->
+	{Commands, State};
+websocket_init(State={Commands, hibernate}) ->
+	{Commands, State, hibernate}.
+
+websocket_handle(_, State) ->
+	{[], State}.
+
+websocket_info(_, State) ->
+	{[], State}.