This commit introduces the BPF maps used by the sessionAffinity feature:

- lb{4,6}_affinity:     used to store current affinity sessions;
                        key = svc id + client_id (either netns cookie or
                        client ip addr), val = last_used and backend_id.

- lb_affinity_match:    used to check whether a backend still belongs
                        to a svc; key = svc id + backend id; the map
                        is going to be shared between ipv4 and ipv6, as
                        svc and backend id pools are the same for ipv4
                        and ipv6.

Signed-off-by: Martynas Pumputis <m@lambda.lt>

First, this commit extends the lb{4,6}_service struct by allowing to
store the fact that sessionAffinity is enabled and a sessionAffinity
timeout (in seconds). The latter is stored only in the placeholder
svc value (slave=0), and it is stored in the backend_id field (thus
the change to the union type).

Second, it adds the provisioning of the sessionAffinity field from k8s
to the BPF maps.

Signed-off-by: Martynas Pumputis <m@lambda.lt>

This commit adds the sessionAffinity feature implementation to
bpf_sock.c and lb.h.

The basic principle is before selecting a svc backend, check whether
there is an active session for the given client. If so, check whether
backend still belongs to the svc.

In the case of bpf_sock.c, the client is identified by a netns cookie,
while in lb.h by the client IP addr (in the former the IP addr is not
available).

Signed-off-by: Martynas Pumputis <m@lambda.lt>

For new deployments it's going to be enabled, as the default in the helm
template is true, and for old - disabled, as cilium-agent defaults to
false.

Signed-off-by: Martynas Pumputis <m@lambda.lt>