[kvlonge]

Why

Sixteen DuckShop staging models output columns that their MODEL headers do not declare: the
source_seq and source_loaded_at load metadata from @sql.load_metadata, and in some models a
payload column such as contents, items or tier_history. Downstream models read those columns,
so a table built from the declared columns alone cannot serve them. Compiling every model against
empty tables created from the declarations failed for 17 downstream models with an invalid
identifier.

Changes

• Declare the missing output columns of 16 staging models (catalogue, growth, inventory, partners,
  storefront and support), with the types and nullability the models produce.
• Declarations only: no model body, test or compiled SQL changes.

Verification

• sqb compile with sqlbuild 0.187.0: 0 errors, 0 warnings.
• The core copy's compiled checksum is unchanged
  (1ff1d0955ea3839b3c969707fd47ca9a4d7d7216753759d23911836f6ea5879d).
• DuckShop pytest suites (benchmarks/duckshop/tests, project/tests): 291 passed, 3 skipped.
• A name check of every core model against tables built from the new declarations resolves every
  column read.
• make check-ci passes.