pulse: fix possible overflow for buffer_time option

The int cast is wrong and unnecessary as internal->buffer_time is a
pa_usec_t (which is uint64_t). It does not even make sense to cast
it to a uint as it being multiplied by the sample rate (and then
divided by 1000000 before it get fitted into battr.tlength, which
is a uint32_t).

Also ditch the "+7" trick for format->bits as the function would
have returned already if it isn't 8/16(/24).

src/plugins/pulse/ao_pulse.c

@@ -258,8 +258,8 @@ int ao_plugin_open(ao_device *device, ao_sample_format *format) {
     /* buffering attributes */
     battr.prebuf = battr.minreq = battr.fragsize = -1;
 
-    battr.tlength = (int)(internal->buffer_time * format->rate) / 1000000 *
-      ((format->bits+7)/8) * device->output_channels;
+    battr.tlength = internal->buffer_time * format->rate / 1000000 *
+      (format->bits / 8) * device->output_channels;
     battr.minreq = battr.tlength/4;
     battr.maxlength = battr.tlength+battr.minreq;