data-cleanup: fix a messy CSV before you load it

There is only one tool, so there is nothing to confuse it with. Its purpose — cleaning a messy CSV and returning the cleaned file plus a change report — is explicit and unmistakable.

The single name 'data-cleanup' is lowercase kebab-case and readable, and there are no sibling names to clash with. It is noun-based rather than the more conventional verb_noun form, a minor deviation but internally consistent.

One tool is thin for a general-purpose server, but this is a deliberate single-purpose paid microservice where the entire scope is CSV cleanup. The count is defensible yet leaves no room for granular operations.

The tool covers the core cleanup lifecycle end to end: input via URL or inline CSV, whitespace/date/email/phone normalization, dedupe, and a change report. Minor gaps exist — no dry-run/preview, no configurable column mapping or selectable dedupe strategy, and no way to fetch the report separately — but agents can work around these.