Cleanup and fixes for nat-pmp. We no longer do the whole 'try 10 times, doubling the delay in a blocking manner' ttempt from the original code; as I note in the comments above where the attempt is made a single time, this leads to about 8 minutes of nonresponsiveness if the router both doesn't support nat-pmp and doesn't send back a response to let the program know. purple_pmp_create_map() and purple_pmp_destroy_map() now return a gboolean indicating success, as the internal data structures of nat-pmp are certainly not needed elsewhere. The #includes are slightly reordered from the cleanup Mark did (thanks, Mark :) ), as with my OS X 10.3 cross-compiler, anyways, route.h needs to come after the sys includes to be properly handled.