libbpf-tools/ksnoop: perf buffer fix unnecessary junk bits output

[linfeng2999]

In output_trace() bpf_perf_event_output determins its even size (trace_len in this function) based on the fact that trace->buf is at the end of struct trace:

trace_len = sizeof(*trace) + trace->buf_len - MAX_TRACE_BUF;

Commit e9f140f0fea0 ("libbpf-tools: Fix memory leaks in ksnoop/gethostlatency") breaks that rule and we will put 2 pointers size more junk bits into perf buffer. Fix this by moving forward the new added tailing pointers to its proper place(commented as readonly segment).

[chenhengqi]

Good catch, thanks.