CP-50108: Use Ipaddr instead of string-based CIDR address handling

[last-genius]

Refactored several string-based functions that handled CIDR addresses to use the Ipaddr module instead.

Expanded test cases, kept the existent exception behaviour.

My first time doing this with modules (in helpers.ml) - not sure if it's in the right direction at all, please do correct me. The idea was to model the common interface for v4 and v6 to avoid code duplication in the logic part. The common interface the library itself offers (in the top-level Ipaddr.) does not have the functionality I would need here. In particular I'm wondering if I'm not exposing this interface into the public unnecessarily and how would I fix that?

==

These changes passed the BST+BVT test suites.

[last-genius]

Addressed the review comments in a fixup commit - will be squashed before merging

[psafont]

I recommend doing fixups without signing them off, it will make the DCO check to fail and make it obvious the PR shouldn't be merged before the fixups are gone

[last-genius]

This CI failure is weird:

File "ocaml/xapi/nm.ml", line 623, characters 53-80:
623 |                               Prefix.address ip_t |> Ipaddr_unix.V6.to_inet_addr
                                                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
Error: Unbound module Ipaddr_unix

I didn't add this in any of the fixups, and it passed before. Did something in xs-opam change in the meantime?

[last-genius]

Got it, transitive dependencies were turned off on master, had to add the submodule to dune as well. CI passes now.

[lindig]

We are no longer using Scanf for parsing - so I think we should not use this exception (I understand the argument for maintaining previous behaviour). I doubt this exception is handled except at the highest level (might want to grep for it). So with could be a failwith which also would be turned into an internal error at the outermost level. I am fine to merge this as is, though.

[last-genius]

Replaced with a failwith