Add scan-xml-control-chars admin CLI

[jcschaff]

Summary

• New `vcell-su scan-xml-control-chars` command (read-only) walks every row in `vc_biomodelxml` and `vc_mathmodelxml`, joining `vc_biomodel`/`vc_mathmodel` and `vc_userinfo` for the model id and owner. Reports each row whose CLOB contains any codepoint rejected by `XmlChars.isValidXml10Char` (C0 controls except TAB/LF/CR, unpaired surrogates, non-character codepoints, and `U+FFFD` per project policy).
• Output is TSV: `kind, model_id, userid, offset, cp_hex, snippet`. Defaults to first-bad-codepoint per row; `--all-occurrences` enumerates every hit.
• Hardened against the failure modes of the previous scratch implementation: `PrintWriter` is autoFlush so progress survives interrupts; status goes to stderr (no log4j2 dependency); CLOBs above `--max-clob-mb` (default 64) are skipped with a noted reason; per-batch progress every 500 rows.

Why

PR #1676 fixed the import-side charset hygiene that was producing the corrupt CLOBs and PR #1677 added input-validation guards so future writes can't reach this state. This PR is the third leg: tell us how widespread the existing corruption is, so we can decide repair-in-place vs. mark-broken without speculation.

Test plan

• `mvn -pl vcell-admin -am compile` succeeds.
• `vcell-su scan-xml-control-chars --help` prints expected options.
• Run against vcell-dev and confirm the output TSV identifies BioModel 311226221 and 311875206 (already known corrupt).
• Spot-check at least one mathmodel hit, if any, to confirm both queries work.

Notes

• Stacked on `xmlchars-validation` (Validate XML chars in name and sbmlName setters #1677) since it imports `org.vcell.util.xml.XmlChars`. Rebase or merge once Validate XML chars in name and sbmlName setters #1677 lands.
• Read-only by design; no DDL, no UPDATE/INSERT/DELETE. Repair tooling is intentionally out of scope.

🤖 Generated with Claude Code