feat(recommendation): For You exploration attribution — taste vs fresh picks

Milestone #127 step 2 (#1249). For You deliberately blends two
populations — a head of top-scored taste picks and a tail sampled from
deeper ranking (the freshness injection) — but the metrics judged it as
one blob, so its skip rate couldn't distinguish "the taste engine is
missing" from "the freshness tax is too high". That number decides the
exploration share before we tune it.

- Migration 0038: nullable pick_kind ('taste'|'fresh') on both
  playlist_tracks (stamped at snapshot build) and play_events (frozen at
  play-ingestion — the snapshot rebuilds daily, so attribution cannot be
  reconstructed at read time).
- Builder: pickHeadAndTail marks head=taste / tail=fresh; the small-pool
  fallback is all taste (top-N-by-score IS the taste mechanism). Other
  variants persist NULL.
- Ingestion: for_you plays (live + offline replay) look the track up in
  the user's current snapshot; not found → unattributed, never guessed.
- Metrics: For You's row gains a breakdown (taste / fresh / earlier
  unattributed plays), parent row stays the sum; web card renders the
  sub-rows indented with the same baseline deltas + low-data dimming.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TsF3cNoKrqCYsU78cXC8U6

internal/playlists/foryou_test.go

@@ -243,3 +243,53 @@ func TestPickTopN_DiversityCap(t *testing.T) {
 		t.Errorf("len = %d, want 3 (artist 100 capped at 3)", len(got))
 	}
 }
+
+func TestPickHeadAndTail_MarksPickKinds(t *testing.T) {
+	// 100-deep pool with headN=20, tailN=5: the 20 head entries are the
+	// taste picks, the 5 tail entries the freshness injection (#1249).
+	in := make([]recommendation.Candidate, 0, 100)
+	for i := 0; i < 100; i++ {
+		in = append(in, makeCand(i+1, i+1, i+1, float64(100-i)))
+	}
+	got := pickHeadAndTail(in, testUserID, "2026-05-07", time.Now(), 20, 5)
+	if len(got) != 25 {
+		t.Fatalf("len = %d, want 25", len(got))
+	}
+	for i := 0; i < 20; i++ {
+		if got[i].PickKind != pickKindTaste {
+			t.Errorf("head[%d].PickKind = %q, want %q", i, got[i].PickKind, pickKindTaste)
+		}
+	}
+	for i := 20; i < 25; i++ {
+		if got[i].PickKind != pickKindFresh {
+			t.Errorf("tail[%d].PickKind = %q, want %q", i-20, got[i].PickKind, pickKindFresh)
+		}
+	}
+}
+
+func TestPickHeadAndTail_SmallPoolAllTaste(t *testing.T) {
+	// The small-pool fallback is pure top-N-by-score — that IS the taste
+	// mechanism, so nothing on this path is an exploration pick.
+	in := []recommendation.Candidate{
+		makeCand(1, 10, 100, 1.0),
+		makeCand(2, 11, 101, 0.9),
+		makeCand(3, 12, 102, 0.8),
+	}
+	got := pickHeadAndTail(in, testUserID, "2026-05-07", time.Now(), 5, 2)
+	for i, rc := range got {
+		if rc.PickKind != pickKindTaste {
+			t.Errorf("got[%d].PickKind = %q, want %q (fallback is all taste)",
+				i, rc.PickKind, pickKindTaste)
+		}
+	}
+}
+
+func TestPickTopN_NoPickKind(t *testing.T) {
+	// Songs-like-X and the discovery mixes don't split; their rows must
+	// persist pick_kind NULL (empty string here).
+	in := []recommendation.Candidate{makeCand(1, 10, 100, 1.0)}
+	got := pickTopN(in, testUserID, "2026-05-07", time.Now(), 25)
+	if len(got) != 1 || got[0].PickKind != "" {
+		t.Errorf("pickTopN PickKind = %q, want empty (persists as NULL)", got[0].PickKind)
+	}
+}