fix: change limits and resources to take into account jvm startup bursts

When a pod with JVM starts, it has some bursts in terms of cpu and memory. While the previous limits were mostly correct for the actual application, it didn't take into account those bursts.

Additionally the requested resources have been increased a bit in order to not trigger autoscale on startup. This is a temporary solution, in future we should investigate native builds to circumvent JVM altogether.