CNF does not lift INC through non-lambda nodes

CNF lifts INC through LAM:

```
λx.↑A  →  ↑(λx.A)
```

But it does **not** lift INC through other constructors. For example:

```
#T{↑1, 2, 3}
```

should reduce to:

```
↑#T{1, 2, 3}
```

but currently doesn't.

Either INC lifting should be generalized to all node types, not just LAM, or LAM should not lift INCs.

This discrepancy is confusing