perf mem-events: Avoid uninitialized read

pmu should be initialized to NULL before perf_pmus__scan loop. Fix and
shrink the scope of pmu at the same time. Issue detected by clang-tidy.

Fixes: 5752c20f3787 ("perf mem: Scan all PMUs instead of just core ones")
Signed-off-by: Ian Rogers <[e-mail]>
Acked-by: Namhyung Kim <[e-mail]>
Cc: Ravi Bangoria <[e-mail]>
Cc: Nick Desaulniers <[e-mail]>
Cc: Yang Jihong <[e-mail]>
Cc: Huacai Chen <[e-mail]>
Cc: Nathan Chancellor <[e-mail]>
Cc: Kan Liang <[e-mail]>
Cc: [e-mail]
Cc: Ming Wang <[e-mail]>
Cc: Tom Rix <[e-mail]>
Cc: [e-mail]
Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Namhyung Kim <[e-mail]>

tools/perf/util/mem-events.c

···
 {
 	int i = *argv_nr, k = 0;
 	struct perf_mem_event *e;
-	struct perf_pmu *pmu;
 
 	for (int j = 0; j < PERF_MEM_EVENTS__MAX; j++) {
 		e = perf_mem_events__ptr(j);
···
 			rec_argv[i++] = "-e";
 			rec_argv[i++] = perf_mem_events__name(j, NULL);
 		} else {
+			struct perf_pmu *pmu = NULL;
+
 			if (!e->supported) {
 				perf_mem_events__print_unsupport_hybrid(e, j);
 				return -1;