ipam: report IP owner of non-default pool IPs in multi-pool IPAM

[tklauser]

Before this change, only IPs from the default pool are reported in cilium status --verbose in the "Allocated addresses" section when using multi-pool IPAM mode. With this change, all IPs from non-default pools are reported as well by making use of the fact that the pool name is prepended to the IP in (*multiPoolManager).dump.

Example:

% ks exec -it cilium-n8xw7 -- cilium status --verbose
[...]
IPAM:                   IPv4: 2 IPAM pool(s) available,
Allocated addresses:
  10.10.0.46 (router)
  10.10.0.48 (kube-system/coredns-5d78c9869d-d4nts [restored])
  10.10.0.61 (kube-system/hubble-relay-597b7bdff8-pwl69 [restored])
  10.10.0.62 (health)
  jupiter/192.168.0.15 (default/nginx-jupiter)
[...]

Note that IPs from non-default pools (in this case the jupiter pool) are reported with the pool name prefixed.

For #25470

[tklauser]

/test

[tklauser]

/test

[gandro]

Thanks for fixing this. Have you also considered changing the type of the IPv4Allocator.Dump() interface to store the pool name separately? To my knowledge, the Allocator interface (where Dump is a part of) while exported is actually not used outside the IPAM package, so I think we should consider changing the signature of ipam.Allocator.Dump to contain the pool name if possible.

The ipam.IPAM.Dump signature can remain as is, I would only change ipam.Allocator.Dump so we can transport the information without having to split strings.

[tklauser]

Thanks for fixing this. Have you also considered changing the type of the IPv4Allocator.Dump() interface to store the pool name separately? To my knowledge, the Allocator interface (where Dump is a part of) while exported is actually not used outside the IPAM package, so I think we should consider changing the signature of ipam.Allocator.Dump to contain the pool name if possible.

The ipam.IPAM.Dump signature can remain as is, I would only change ipam.Allocator.Dump so we can transport the information without having to split strings.

Thanks for the suggestion. I haven't considered this so far but it would make the interface a lot cleaner. I'd probably defer that change to a follow-up PR, so this rather small PR can be backported (given it adds additional debugging/tourbleshooting information for multi-pool IPAM) without much internal API refactoring.