feat(person): add nullable generation column to persons (#689)

Flyway V70: SMALLINT generation column with CHECK(0..10) and partial
index over non-null rows. Person.generation field surfaces it through
the JPA model. Pre-import rows and persons outside the curated family
graph legitimately stay null; the canonical importer (next commits)
back-fills via preferHuman so a human-edited value is never lost.

Refs #689

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

backend/src/main/resources/db/migration/V70__add_generation_to_persons.sql

@@ -0,0 +1,24 @@
+-- #689: persist the hand-curated "G 0…G 5" generation index from
+-- canonical-persons.xlsx so the Stammbaum layout can use it as a strict
+-- rank anchor (replacing the current iterative longest-path heuristic that
+-- silently misplaces loose spouses with their own parents in the graph).
+--
+-- Nullable: pre-import rows and persons outside the curated family graph
+-- legitimately have no generation. The canonical importer back-fills via
+-- preferHuman on the next run; a human-edited value is never overwritten
+-- (see ADR-025).
+
+ALTER TABLE persons ADD COLUMN generation SMALLINT;
+
+-- Allowlist of valid generation indices. The upper bound is intentionally
+-- loose — current data tops out at G 5, but a future G 6 → G 10 widening
+-- needs no migration. A G −1 ancestor would require a separate one-shot
+-- shift migration (out of scope here; the layout's normalise step already
+-- handles negative seeds at render time).
+ALTER TABLE persons ADD CONSTRAINT chk_generation_range
+    CHECK (generation IS NULL OR generation BETWEEN 0 AND 10);
+
+-- Partial index: only the curated rows (≈ 163 of 1,105) ever get a value,
+-- and the layout only ever queries for non-null rows.
+CREATE INDEX idx_persons_generation ON persons (generation)
+    WHERE generation IS NOT NULL;