Fix thread safety issue on clearing cache

Historically, we havent held a lock on ForceClearCache and
ClearCacheIfNecessary. This is not thread safe because kAllPullAtomInfo
must be accessed in a lock, especially now that pullers can be
registered/unregistered.

Test: atest statsd_test, wrote a repro cl.
Bug: 168156854
Bug: 173552790
Change-Id: I47d53a6d9d274bca4c78dbfd87e0097091b7b8cb
Merged-In: I47d53a6d9d274bca4c78dbfd87e0097091b7b8cb
Merged-In: I8bee7a0a6acecc1274d5acc0adb44c5dde8862e4
(cherry picked from commit f9a4bb18ffa1605cced3ee4b161675a0c88941be)
(cherry picked from commit db097214b109be7ca1b3ae470d4a8354c98a5b62)

diff --git a/cmds/statsd/src/external/StatsPullerManager.cpp b/cmds/statsd/src/external/StatsPullerManager.cpp
index 8a9ec74..99c39f6 100644
--- a/cmds/statsd/src/external/StatsPullerManager.cpp
+++ b/cmds/statsd/src/external/StatsPullerManager.cpp

@@ -334,6 +334,7 @@
 }
 
 int StatsPullerManager::ForceClearPullerCache() {
+    std::lock_guard<std::mutex> _l(mLock);
     int totalCleared = 0;
     for (const auto& pulledAtom : kAllPullAtomInfo) {
         totalCleared += pulledAtom.second->ForceClearCache();
@@ -342,6 +343,7 @@
 }
 
 int StatsPullerManager::ClearPullerCacheIfNecessary(int64_t timestampNs) {
+    std::lock_guard<std::mutex> _l(mLock);
     int totalCleared = 0;
     for (const auto& pulledAtom : kAllPullAtomInfo) {
         totalCleared += pulledAtom.second->ClearCacheIfNecessary(timestampNs);