Add Memory metric to HPA policy

[rtclauss]

Related to #13.

When doing load testing of the entire solution, including trader, I notice that the trader pod is occasionally killed due to OOM, OOMKilled. The k8s runtime is killing the pod due to the pod more than the maximum amount of memory allowed.

Although adding a memory metric to HPA policy might mask the underlying issue in trader it could be nice to have this as another scaling policy in our toolbox.