Fix spurious has-type error when unpacking a TypeVarTuple with a constrained TypeVar

[omkar-334]

Fixes #21693.

Unpacking a TypeVarTuple variable like y = (*result,) gave a bogus Cannot determine type of "result" error, but only when the enclosing function also had a value-constrained TypeVar. The result silently became tuple[Any, ...].

The constrained TypeVar is what makes mypy re-check the body once per value, deep-copying the AST through TransformVisitor. Every visitor there recurses into its children and remaps Var references so they stay consistent within a copy, except visit_star_expr, which returned StarExpr(node.expr) without touching the inner expression. So in the copy the assignment target result got a fresh Var while *result still pointed at the original, which is never assigned, so its type never resolves.

The fix is a one-liner: recurse into the inner expression like everything else. I also carry over the valid flag, which the old code was dropping.

Added a regression test that fails on master and passes here. Full check, semanal, fine-grained, and transform suites pass.

[github-actions]

According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅