Fit-score the whole catalog against one dataset

Score every served curated template against one target dataset in a single call — the gallery's hot path once a dataset is in context. Loads the target profile ONCE, then loops the served catalog reusing each entry's cached spec, running the SAME resolver as apply (fit and apply never disagree). Returns a cards-lean fit per template: grade (great|good|partial|poor), reasons (always present on a non-great grade — never a bare score), the matched/total microcopy and required_met. Read-only, no writes, no bake; the profile build + scan run off the event loop. Auth: the v3 workspace gate + the target dataview ACL (mirrors resolve-mapping). table_item_id is optional — the dataview's latest data item is used when omitted.