some bug reports for subscr usage

hi, i found some leaking on the subscriber ressource usage. this patch is not final. maybe we find better solutions. 1. trans_alloc() may not allocate subscriber for transaction, because may be already done before, so i removed subscr_get() from transaction.c -> on outgoing calls we allocate subscriber ressource (by searching for it) and use it for transaction. -> in incomming calls we use the subscriber pointer from the lchan. 2. on an incomming call/sms, the lchan holds a subscriber ressource, so i added subscr_get() after successfully calling trans_alloc() 3. when _send_sms_buffer() is called, we already hold a subscriber ressource for the receiver, so the subsequently called function sms_from_text() must not allocate the receiver again. if this function fails, the allocated subcriber ressource must be released in _send_sms_buffer(). 4. on trans_free we don't need to remove pending paging processes, because it is done when paging expires. anyway if we want to keep it in trans_free(), we must remove paging_remove_request() from paging_T3113_expired(), because it will then be done when transaction is removed. also we need to add paging_remove_request() when sms is removed due to paging expiery. i also fixed linux call router for compiling with openbsc without any patch at all. by linking the source directory to the lcr source directory, the "configure" script automatically detects it and it will compile. (latest git) regards, andreas