[Plasma] Jemalloc incorrectly reports memory usage after upgrade

Steps to Repro:

1) Create a 2 node cluster - kv-(index,n1ql)

2) Load travel-sample

3) Wait for data and indexes to load.

Check http://<indexer_ip>:9102/stats.

There is huge difference between memory_used and memory_used_storage.

Because of this "index data size" and "Max Indexer Ram used" do not match.

Indexer RAM allocated : 512 MB

index data size: 3.92MB 

Max Index RAM Used %: 48.1 
 

This issue is also seen on 5.0.0-3217.

We noticed this Issue in FDB->Plasma Upgrade. See below for Logs.

Before Upgrade:

https://s3.amazonaws.com/bugdb/jira/Before_Upgrade/collectinfo-2017-09-12T110813-ns_1%40172.16.12.9.zip

After Upgrade:

https://s3.amazonaws.com/bugdb/jira/after_upgrade/collectinfo-2017-09-12T111640-ns_1%40172.16.12.9.zip

10 mins after After Upgrade:

https://s3.amazonaws.com/bugdb/jira/10_mins_after_upgrade/collectinfo-2017-09-12T112433-ns_1%40172.16.12.9.zip

20 mins after After Upgrade:

https://s3.amazonaws.com/bugdb/jira/20_mins_after_upgrade/collectinfo-2017-09-12T113333-ns_1%40172.16.12.9.zip