river-control: send output on success, fix issues

Add an output arg to the success event on the callback. This allows for
implementing commands that return values, which are planned.

Replace the array of null terminated strings with a series of requests
each adding a single string to the args array. This is more idiomatic
wayland.

Add a seat argument to the run_command request to allow for proper
multi-seat support in the future.

Add missing destructor request.

protocol/river-control-unstable-v1.xml

@@ -19,32 +19,57 @@
   <interface name="zriver_control_v1" version="1">
     <description summary="run compositor commands">
       This interface allows clients to run compositor commands and receive a
-      success/failure response with a failure message if needed.
+      success/failure response with output or a failure message respectively.
+
+      Each command is built up in a series of add_argument requests and
+      executed with a run_command request. The first argument is the command
+      to be run.
+
+      A complete list of commands should be made available in the man page of
+      the compositor.
     </description>
 
-    <request name="run_command">
+    <request name="destroy" type="destructor">
-      <description summary="run a compositor command">
+      <description summary="destroy the river_control object">
-        A complete list of commands will be found in the man page for the
+        This request indicates that the client will not use the
-        compositor.
+        river_control object any more. Objects that have been created
+        through this instance are not affected.
       </description>
-      <arg name="command" type="array" summary="the command to run as a series
+    </request>
-        of null-terminated strings"/>
+
+    <request name="add_argument">
+      <description summary="add an argument to the current command">
+        Arguments are stored by the server in the order they were sent until
+        the run_command request is made.
+      </description>
+      <arg name="argument" type="string" summary="the argument to add"/>
+    </request>
+
+    <request name="run_command">
+      <description summary="run the current command">
+        Execute the command built up using the add_argument request for the
+        given seat.
+      </description>
+      <arg name="seat" type="object" interface="wl_seat"/>
       <arg name="callback" type="new_id" interface="zriver_command_callback_v1"
-        summary="callback object to receive success/error events"/>
+        summary="callback object"/>
     </request>
   </interface>
 
   <interface name="zriver_command_callback_v1" version="1">
     <description summary="callback object">
       This object is created by the run_command request. Exactly one of the
-      success or failure events will be sent.
+      success or failure events will be sent. This object will be destroyed
+      by the compositor after one of the events is sent.
     </description>
 
     <event name="success">
       <description summary="command successful">
-        Send when the command has been successfully received and validated by
+        Sent when the command has been successfully received and executed by
-        the server and will be carried out.
+        the compositor. Some commands may produce output, in which case the
+        output argument will be a non-empty string.
       </description>
+      <arg name="output" type="string" summary="the output of the command"/>
     </event>
 
     <event name="failure">