feature: TensorTrain <-> TreeTN conversion API

[shinaoka]

Description

There is no public API to convert between TensorTrain<Complex64> (simplett) and TreeTN<TensorDynLen, usize> (treetn). This is the single largest usability barrier for the quanticstransform crate.

Every user who wants to apply a quantics operator must write ~100 lines of boilerplate to:

1. Convert TensorTrain to TreeTN (construct TensorDynLen with column-major layout, handle boundary tensors, create Index objects)
2. Remap site indices to match operator input indices via get_input_mapping
3. Apply the operator
4. Extract results from the output TreeTN

The integration test (crates/tensor4all-quanticstransform/tests/integration_test.rs) contains helper functions (tensortrain_to_treetn, contract_treetn_to_vector) that are copy-pasted by every user.

Additional gaps

• No TensorTrain<f64> -> TensorTrain<Complex64> conversion (TCI produces f64, operators require Complex64)
• No QuanticsOperator re-export from quanticstransform lib.rs

Suggested API

// In tensor4all-treetn or a bridging crate:
pub fn tensortrain_to_treetn(tt: &TensorTrain<Complex64>) -> (TreeTN<TensorDynLen, usize>, Vec<Index>);

// High-level convenience:
pub fn apply_operator_to_tensortrain(op: &QuanticsOperator, tt: &TensorTrain<Complex64>) -> Result<TensorTrain<Complex64>>;

Found by

Reported by 5 out of 6 automated test personas as the primary friction point.