fix(qdp): apply shared max qubit validation

[viiccwen]

Related Issues

Closes #1262

Changes

• Bug fix
• New feature
• Refactoring
• Documentation
• Test
• CI/CD pipeline
• Other

Why

Several amplitude f32 batch and GPU-pointer paths calculated 1 << num_qubits without first using QDP's shared validate_qubit_count function. Excessive inputs could therefore panic on shift overflow or proceed toward an impractical allocation instead of returning the established MahoutError::InvalidInput. QDP also retained an unused duplicate MAX_QUBITS macro in the CUDA kernel configuration.

How

• Apply validate_qubit_count before state-length calculations in the uncovered float32 batch and float32/float64 GPU-pointer single and batch paths.
• Keep the Rust MAX_QUBITS constant as the single effective validation policy.
• Remove the unused CUDA MAX_QUBITS definition and update the stale Rust comment.
• Extend the existing maximum-qubit and GPU-pointer test modules with regression coverage for all affected entry points.

Runtime device-capacity planning remains outside this cleanup's scope.

Checklist

• Added or updated unit tests for all changes
• Added or updated documentation for all changes