docs/troubleshooting.md describes CRD auto-registration that doesn't exist

[phaedrus1992]

Found while reviewing #556 — pre-existing, unrelated to that PR.

docs/troubleshooting.md:53 recommends granting
customresourcedefinitions.apiextensions.k8s.io -- get, create, patch -- "for CRD
self-registration", and a few lines below says "The operator auto-registers the CRD on startup, but
if RBAC blocks that, install it manually."

There's no such code. crates/servarr-operator/src/main.rs's only apiextensions-adjacent behavior is
the servarr-operator crd subcommand, which prints the CRD YAML to stdout for the user to
kubectl apply themselves (main.rs, Commands::Crd => { controller::print_crd()?; ... }) — it
never touches the cluster's apiextensions.k8s.io API. Grepping the whole operator crate for
apiextensions/CustomResourceDefinition outside that doc turns up nothing that creates or patches a
CRD.

So either this was aspirational documentation for a feature that was never built, or it described a
feature that was since removed. Either way the RBAC guidance is currently wrong twice over: it
recommends create/patch on a cluster-scoped resource the operator never writes to, and the
"auto-registers on startup" troubleshooting step sends someone chasing a problem that doesn't exist.

Fix: either implement CRD auto-registration for real (get+create+patch would then be justified), or
strike both the "auto-registers" troubleshooting step and the create/patch verbs from the doc, leaving
just the manual servarr-operator crd | kubectl apply -f - path that already exists a few lines below.