Improve malloc implementation

[mprobson]

Original issue: https://charm.cs.illinois.edu/redmine/issues/971

---

From core discussion:

• Should we lift the ugni layer's implementation of mempool for use everywhere?
• PAMI layer's usage of malloc should also be looked over as a way to generally improve malloc
• In general, we want to make sure that we're using the most optimal malloc implementation on every machine

The first step is to do some analysis and gather consensus on a path forward