fix(dbt): write column marker to config.meta so dbt Fusion (v2) accepts the YAML

[FredrikBakken]

dbt sync stamps every provenance block under config.meta.datacontract_cli - model entries, test entries and version bullets - except the per-column "generated" marker, which was written to a top-level meta: on the column.

dbt-core (v1) accepts that (deprecated), but dbt Fusion (v2) validates model YAML against a closed schema and rejects it outright:

Property meta is not allowed. (yaml-schema: dbt YAML files 513)

The marker is only stamped on columns the CLI creates, so the failure was invisible when the model YAML already existed and only surfaced when dbt sync generated the file from scratch - the two merge routines had silently diverged.

config.meta validates under both engines, so write there instead:

• _mark_managed_column writes config.meta.datacontract_cli.generated.
• _is_managed_column reads config.meta and falls back to the legacy top-level location, so projects synced by an older CLI stay recognized (otherwise cleanup/--prune would orphan the columns it used to own).
• _migrate_legacy_column_meta relocates legacy blocks on the next sync, wired into both _merge_model_entry and _merge_versioned_model_entry. User keys sitting alongside ours under meta: are left untouched.

Verified end to end against a dbt project: dbt-fusion dbt parse went from 10 errors to clean, dbt-core 1.12.4 dbt parse stays clean, and dbt ls --select config.meta.datacontract_cli.include_in_tests:true still matches every managed test under v1 (dbt-core back-populates columns.*.meta from config.meta, so nothing downstream regresses).

• Tests pass (uv run pytest)
• Code formatted (uv run ruff check --fix && uv run ruff format)
• Docs updated (if relevant)
• CHANGELOG.md entry added