fix: case-insensitive dead-vehicle matching for TSS scoreboard

Spectra sends events.kills[].offended_unit with uppercase roman
numerals (V/VI) while players[].units[].unit uses lowercase (v/vi).
85% of TSS games are affected — dead-vehicle strikethrough on the
Discord scoreboard never triggered because  was
case-sensitive.

Normalize both sides to lowercase before comparing.

BOT/transform.py

@@ -60,14 +60,14 @@ def _build_units(units: list[dict[str, Any]], translate, dead: set[str] | None =
     Prefer an explicit per-unit ``dead``/``died`` flag if Spectra provides one; otherwise
     fall back to the ``dead`` set cross-referenced from ``events.kills`` (see ``_dead_units_by_uid``).
     """
-    dead = dead or set()
+    dead_lc = {s.lower() for s in (dead or set())}
     out: list[dict[str, Any]] = []
     for u in units or []:
         internal = strip_model_prefix(u.get("unit"))
         if not internal:
             continue
         flag = u.get("dead", u.get("died"))
-        is_dead = bool(flag) if flag is not None else internal in dead
+        is_dead = bool(flag) if flag is not None else internal.lower() in dead_lc
         out.append({
             "internal": internal,
             "name": translate(internal) or u.get("unit_normalized") or internal,
@@ -88,7 +88,7 @@ def _dead_units_by_uid(game: dict[str, Any]) -> dict[str, set[str]]:
         uid = str(k.get("offended_uid") or "")
         unit = strip_model_prefix(k.get("offended_unit"))
         if uid and unit:
-            out.setdefault(uid, set()).add(unit)
+            out.setdefault(uid, set()).add(unit.lower())
     return out