Inventory Import: Items Made On Demand, Unrecognized SKUs Don't Sink The File

The inventory CSV import no longer needs an inventory item to already exist for every SKU. If a row references a product that doesn't have an inventory item yet and has stock to record, the item is created on demand and the stock is adjusted. Zero-stock rows for SKUs with no item are skipped — there's no point creating an empty record.

Rows whose SKU has no matching product at all used to fail the entire import. Now they're skipped and every skipped SKU is listed in the import summary, so the rest of the file goes through and you can fix the unrecognized SKUs afterwards. Rows with real data errors (bad warehouse, bad location, bad stock value) still fail the import — genuine CSV problems are not silently swallowed.

This unblocks operators whose products were created in bulk without ever provisioning inventory items.