Compound assignment to a DECLARED static class field is silently dropped (`K.n += 1` leaves `K.n` unchanged; `=` and `++` work) — both modes

[proggeramlug]

class K { static n = 1; }
K.n = 5;   console.log(K.n);   // 5      (correct)
K.n += 1;  console.log(K.n);   // node 6   Perry 5
K["n"] += 1; console.log(K.n); // node 7   Perry 5
const k = "n"; K[k] += 1; console.log(K.n); // node 8   Perry 5
K.n++;     console.log(K.n);   // node 9   Perry 6   (`++` itself is applied; the three `+=` were lost)

Identical in an ESM .ts module (strict) and a .cts script (sloppy). Compound assignment to a declared static field is silently dropped; plain = (Expr::PutValueSet) and ++ (Expr::PropertyUpdate) on the same field are applied. An undeclared static (class D {}; (D as any).m = 1; (D as any).m += 1 → 2) is correct in both modes, so the dynamic class-property bag is not the problem — the declared static field has its own storage that the = and ++ lanes reach and the Expr::PropertySet lane (o.x += 1, o["x"] += 1, o[k] += 1, and presumably for (K.n of …) / [K.n] = …) does not: the store presumably lands in the class's dynamic-property bag and the next read of K.n comes from the static slot.

Reproduced on fix/propertyset-strict (PR #9519's head, main + two commits that do not touch static fields); the strict lane there is byte-identical to main, so this is pre-existing on main in both modes. Found while building the #9495 fixture (which now uses an undeclared static as its class-ref control instead). Not a strictness or prototype-walk defect — #9459 / #9495 leave it as it was.