change javascript and confernece to use cached speech handles

This is for better performace and to take atvantage of mrcp connections.
There should be no apparent change in usage.
The conference now supports changing the voice via the text string
by starting a text blurb with #voice# eg

#william#This will sound like William.




git-svn-id: http://svn.freeswitch.org/svn/freeswitch/trunk@5664 d0543943-73ff-0310-b7d9-9358b9ac24b2

src/mod/languages/mod_spidermonkey/mod_spidermonkey.h

@@ -122,6 +122,11 @@ struct sm_module_interface {
 typedef struct sm_module_interface sm_module_interface_t;
 typedef switch_status_t (*spidermonkey_init_t) (const sm_module_interface_t ** module_interface);
 
+struct js_session_speech {
+	switch_speech_handle_t sh;
+	switch_codec_t codec;
+};
+
 struct js_session {
 	switch_core_session_t *session;
 	JSContext *cx;
@@ -141,6 +146,7 @@ struct js_session {
 	char *rdnis;
 	char *context;
 	char *username;
+	struct js_session_speech *speech;
 };
 
 JSBool DEFAULT_SET_PROPERTY(JSContext * cx, JSObject *obj, jsval id, jsval *vp)