feat: support loadBalancerClass-scoped service handling

[Skaronator]

Disclaimer! The rust code was basically completely AI written, since I have no clue about Rust. If you don't like specific implementation details then please go ahead and push changes directly, thanks.

I did an end-2-end test run on my Kubernetes Cluster, and the changes work fine. No more conflicts with my MetalLB LoadBalancer! 🎉

---

I wasn't able to build the docker image due to a glibc version mismatch in the build stage (rust:latest) and the run stage (redhat/ubi9-micro:latest). This is most likely that rust:latest now uses debian trixie which ships which a much newer version than redhat/ubi9 uses.

I purpose that we just switch the run stage to trixie as well. Additionally I made the container rootless :)

This is now similar how Vaultwarden Docker is built.

Sidenote: I tried switching to ubi9 in the build stage, but this isn't that simple and redhat is like 5 rust versions behind. So I don't think it worth it.

---

Closes #153
Fixes #150

[korewaChino]

Disclaimer! The rust code was basically completely AI written

As long as you can attest it works properly and doesn't cause issues (and you actually vetted its output for weird stuff), it doesn't bother me.

Also, did you test this against an old config? Would this work when migrating from an older version (as in, no config option was set)

[korewaChino]

hmm... lgtm?

@lleyton I'd like a second opinion before we merge this

[nothingneko]

fully AI, gotta check with the lawyers on this one

[nothingneko]

okay so, AI can't hold copyright, meaning that this code is public domain. that means that this public domain code is incompatible with the GPL (which chisel is licensed under)

if you wanna try again without AI you're more than welcome to, but we have to decline this PR right now. sorry

edit: lawyers and engineers are fighting, hold on

edit 2: engineers won

[korewaChino]

since we can't merge this PR due to it being AI-generated (even though it's outputting somewhat solid code), we'll be closing it for now :/

[nothingneko]

sorry for difficulty with this one, we're still figuring out how AI usage like this should be handled. thank you for the contribution

[korewaChino]

on further deliberation, this actually looks good enough. I'll be merging this. Sorry for the confusion.

Just note that AI-generated code is kind of muddy licensing-wise, but on the merit of the code quality itself, this PR is viable for merge.