Parsing or memory limits and requests is failing for some numbers

It seems that for some numbers, our parsing of the memory requests and limit doesn't work properly. For example `500Mi` or `1.1Gi` will cause problems either in the CO it self directly or in Kubernetes when trying to set them in the statefulsets.

More details can be found in [Strimzi#1435|https://github.com/strimzi/strimzi-kafka-operator/issues/1435).