refactor: relationship model to column lists for compound FK (ADR-0043)

Move the FK column fields String->Vec<String> through all six
layers (AddRelationship/SqlForeignKey AST, RelationshipSchema,
metadata, project.yaml, ReadForeignKey, RelationshipEnd). Metadata
stores comma-joined lists in the existing TEXT cells; project.yaml
endpoints now columns: [a, b] (house style). Executor logic is
multi-column ready: resolve_fk_parent_columns (full-PK F-A +
auto-expand F-D), per-pair type-compat, schema_to_ddl multi-column
emission, pragma FK read grouped by id, auto-name + --create-fk
per-column, multi-column teaching echo. Single-column behaviour
preserved (one-element vecs); all 2181 tests green. The grammar to
parse multi-column input lands next.

docs/adr/0043-compound-pk-foreign-key-references.md

@@ -159,9 +159,14 @@ convention `project.yaml` already uses for `primary_key` and index
 - **Metadata** (`__rdbms_playground_relationships`): no
   `CREATE TABLE` change (the `TEXT` columns and
   `PRIMARY KEY (child_table, child_column)` are untouched).
-  `parent_column` / `child_column` store the list as a JSON array
-  string — uniformly, including `["id"]` for a single column
-  (SQLite has no array type, so a text cell is where a list lives).
+  `parent_column` / `child_column` store the list **comma-joined**
+  in the same text cell (`a,b`; a single column is just its bare
+  name). *As-built note:* the ADR first said "JSON array"; the
+  implementation uses a comma delimiter, which is safe because
+  column identifiers are `[A-Za-z0-9_]+` (no commas — `parser.rs`)
+  and simpler (no `serde_json` dependency). This is an internal
+  encoding detail below fork F-B — the user-visible `project.yaml`
+  is still the `columns: [a, b]` list.
   The actual enforced FK lives on the rebuilt child table's DDL
   (`FOREIGN KEY (a, b) REFERENCES P(x, y)`), emitted by
   `schema_to_ddl`, exactly as the single-column FK is today via the