aya: perf_map: fix bug when max_entries=0

When a perf map has max_entries=0, max_entries is dynamically set at
load time to the number of possible cpus as reported by
/sys/devices/system/cpu/possible.

This change fixes a bug where instead of setting max_entries to the
number of possible cpus, we were setting it to the cpu index of the last
possible cpu.

aya/src/bpf.rs

@@ -117,13 +117,12 @@ impl Bpf {
         for (_, mut obj) in obj.maps.drain() {
             if obj.def.map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY as u32 && obj.def.max_entries == 0
             {
-                obj.def.max_entries = *possible_cpus()
+                obj.def.max_entries = possible_cpus()
                     .map_err(|error| BpfError::FileError {
                         path: PathBuf::from(POSSIBLE_CPUS),
                         error,
                     })?
-                    .last()
-                    .unwrap_or(&0);
+                    .len() as u32;
             }
             let mut map = Map { obj, fd: None };
             let fd = map.create()?;