fix(nyc-taxi): stop tag/term attachments overwriting each other; create owner corpGroup; honest dry-run

[simpleciki]

Fixes #218. Also fixes the Windows Unicode crash reported in #211 for this script.

Four fixes in datasets/nyc-taxi/add_metadata.py:

1. Per-table accumulation — attach_tags/attach_glossary now build the full association list per table and emit one aspect per table, so attachments no longer overwrite each other (previously each emit replaced the whole globalTags/glossaryTerms aspect with a single association — last write wins, and every table ended up with only pipeline_stage). Verified on a quickstart instance (server v1.5.0.6): end state matches the README's Metadata section exactly — raw_trips/staging_trips carry 4 tags + 2 terms, mart carries daily_refresh + pipeline_stage and Freshness SLA + Empty Load + Pipeline Stage.
2. Create the owner group — new create_groups() emits corpGroupInfo for data_platform_team before ownership attachment; the owner chip in the UI now resolves instead of "Not Found".
3. Honest dry-run — --dry-run now prints the planned tag/term/ownership operations and ends with "DRY RUN — no changes made" instead of the unconditional "✅ Metadata complete" banner (matching what add_lineage.py --dry-run already does).
4. Windows-safe output — the ✓/→ characters crash default cp1252 Windows consoles (Sample dataset scripts crash on Windows (cp1252): UnicodeEncodeError printing check marks #211, reported by another hackathon participant); a two-line sys.stdout.reconfigure at startup fixes it for this script (tested on a native cp1252 console, exit 0). The same two-liner applies to the other dataset scripts if you'd like a follow-up.

No changes to lineage, ingestion recipes, or the .db files.