CNS in-mem cache to use containerid as key if manageendpointstate is enabled

[tamilmani1989]

Reason for Change:

This PR fixes duplicate ip assignment to pods if cns looksup/updates podinfo cache based on pod name and namespace. This PR updates to lookup based on containerid instead if managedendpointstate is set.

• containerd invokes cni with containerid say "c1" and cni allocates ip from cns and returns success but something failed in containerd and it issues add with new container id say "c2" for same pod before calling delete with "c1"
• cni receives add for same pod with "c2" containerid and request cns to allocate ip and cns returns an available ip and marks it allocated
• cni receives del for same pod with "c1" containerid and request cns to release ip allocated for "c1" container id but since cns keeps pod name + namespace as key in-memory cache it may release ip allocated for containerid "c2"
• cni receives add for some other pod and cns may endup assigning same ip to this pod but cilium agent has a check to catch these duplicate ip assignment and throw error and cni add call will be retried

Issue Fixed:

Requirements:

• uses conventional commit messages
• includes documentation
• adds unit tests

Notes: