Instead we now perform a partial transformation which we can then later extend.
This means we no longer need to rely on the previous instruction (prior to startPos) being already inlined.
This way the IndexRangeTransform also works without the duplicate StatementTransform pass.