fix(csv): promote "Output CSV column order" out of the Metadata section

[JakeSCahill]

Problem

The csv input's Description() nests === Output CSV column order — a how-to for producing deterministic CSV output (prose + a Bloblang escape_csv example) — as a level-3 subsection under == Metadata. It isn't metadata.

Downstream tooling in the Redpanda Connect docs extracts the == Metadata section verbatim into a regenerated metadata reference partial, and because a level-3 heading doesn't terminate the section, the entire "Output CSV column order" subsection (including the YAML block) is swept into the generated metadata docs. See the analysis in redpanda-data/rp-connect-docs#459.

Fix

Promote the heading from === Output CSV column order to a sibling == Output CSV column order section. It's still on the csv input page, just no longer nested inside — and no longer part of the Metadata block.

Note

Two connectors were found doing this; the other (gcp_cloud_storage, === Credentials) lives in redpanda-data/connect and is fixed there separately. A legitimate use of === under Metadata (grouping fields by operation, e.g. the nats_kv processor) is intentionally left as-is.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}