fix: scope LoadBalancer lookup to IpfsNode instance and kind

[plsgiveup]

Summary

• Scope the existing LoadBalancer Service lookup to the IpfsNode instance and node kind with one combined label selector.
• Add a regression test for the Kubernetes list request generated for storage and bootstrap nodes.

Why

ListParams::labels() replaces the prior selector. The second call discarded the instance filter and always selected storage Services, so Services belonging to another IpfsNode could be counted. A new node could then skip creating its own LoadBalancer Service and wait for an external IP that never arrives.

Validation

• cargo +nightly fmt --all -- --check
• cargo +nightly check --locked --offline --all-targets --all-features
• cargo +nightly clippy --locked --offline --all-targets --all-features -- -D warnings
• cargo +nightly test --locked --offline --all-targets --all-features (1 active test passed; 2 existing telemetry tests remain ignored because they require a trace exporter)
• Built both ipfs-operator and bootstrap-web Containerfile targets.
• Ran the controller in a local kind v1.37 cluster with MetalLB. One bootstrap node, two single-replica storage nodes, and a two-replica storage node produced five correctly labeled LoadBalancer Services with distinct external IPs and five Ready IPFS pods. A repeat reconcile preserved the Service UIDs.