Handle ParameterExpression in qiskit2tq conversion

[yurekami]

Summary

• Add support for converting Qiskit circuits with unbound parameters (ParameterExpression) to TorchQuantum
• Previously, attempting to convert a circuit with symbolic parameters would fail with: TypeError: ParameterExpression with unbound parameters cannot be cast to a float.

Changes

• Add _has_unbound_params() helper to detect unbound ParameterExpression objects
• Add _convert_params() helper to safely convert gate parameters
• Update qiskit2tq_op_history() to handle unbound parameters
• Update qiskit2tq_Operator() to handle unbound parameters
• Remove debug print statement from qiskit2tq_op_history()

Behavior

When a circuit has unbound parameters, the converted TorchQuantum layer will have trainable gates without initial parameter values (init_params=None), allowing them to be trained during optimization.

Example

from qiskit.circuit.library import RealAmplitudes
from torchquantum.plugin import qiskit2tq

# Previously failed with TypeError
ansatz = RealAmplitudes(2, reps=1)
tq_layer = qiskit2tq(ansatz)  # Now works!

Test plan

• Convert RealAmplitudes circuit with unbound parameters
• Convert circuit with ParameterVector
• Verify bound parameters still work correctly
• Verify mixed circuits (fixed + parameterized gates) work

Fixes #198

🤖 Generated with Claude Code