docs(operator): singleton has no default PDB; recommend node auto-repair

[sklarsa]

Draft: hold until the operator release that contains questdb/questdb-enterprise-operator#448. Until then the published behaviour is still the old one.

Operator change: a one-instance cluster no longer gets a PodDisruptionBudget by default, so a node drain or node upgrade restarts the database instead of hanging.

• known-limitations.md: "Singleton drain is blocked by default" becomes "Draining a singleton's node restarts the database".
• configuration.md: the default PDB applies to replicated clusters only; how to opt a singleton back in.
• operations/database.md: the node maintenance section is rewritten for the new default.
• Adds a note recommending node auto-repair for a node that is NotReady while its machine keeps running (the other half of questdb/questdb-enterprise-operator#64).

Before merging:

• reference/api.md is not touched; it is regenerated from the operator at release and will pick up the new field text then.
• This overlaps docs(operator): add safe node-pool maintenance runbook #557, which edits the same opening of the node maintenance section in operations/database.md. Whichever merges second needs a rebase.
• The auto-repair sentence says AKS repairs unhealthy nodes automatically and that EKS needs node auto repair enabled on the managed node group. Worth a check against current cloud docs.

🤖 Generated with Claude Code

[github-actions]

🚀 Build success!

Latest successful preview: https://preview-581--questdb-documentation.netlify.app/docs/

Commit SHA: e429923

📦 Build generates a preview & updates the link on each commit.