Fix nested control-flow padding state

[sankalpsthakur]

Summary

Fix nested control-flow padding so traversal state stays scoped to the block being visited and rebuilt control-flow operations retain their scheduled block index.

Details and comments

BlockBasePadder reused _idle_after and _current_block_idx while recursively visiting nested blocks. A nested visit could therefore overwrite the parent block's timing state. The control-flow rebuild path also dropped the scheduled block index, and the final fast-path barrier call passed its time and block arguments in the wrong order.

This change preserves and restores parent traversal state, carries the scheduled index through rebuilt control-flow operations, and corrects the barrier arguments. Focused regressions cover nested branch traversal and the final fast-path barrier.

Validation:

• pytest -q test/unit/transpiler/passes/scheduling/test_scheduler.py test/unit/transpiler/passes/scheduling/test_dynamical_decoupling.py (159 passed)
• Ruff lint and format checks
• mypy on the changed scheduling and test modules
• Towncrier draft rendering
• git diff --check

Fixes #3077

This is the focused control-flow metadata subset originally catalogued in #3005; the separate idle-qubit detection bug in #3005 remains outside this PR.

Why this matters

Corrupted scheduling metadata in dynamic circuits can shift control-flow start times and attach barriers to the wrong block. That breaks downstream padding and decoupling passes silently — the kind of bug that only surfaces on nested if_test/while circuits.

Related

• Bug report (this PR closes): BlockBasePadder corrupts scheduling metadata for rebuilt control-flow operations #3077
• Broader tracker (partial): Multiple control-flow bugs in block_base_padder.py #3005 — idle-qubit detection remains open

AI/LLM disclosure

• I didn't use LLM tooling, or only used it privately.
• I used OpenAI Codex and Cursor to help write this PR description.
• I used OpenAI Codex and Cursor to generate or modify code, tests, and release-note documentation.

[TsafrirA]

@sankalpsthakur could you open an issue with a bug report explaining what's the issue you're encountering?

[sankalpsthakur]

Opened #3077 with a runnable reproduction and scoped it to the control-flow metadata bugs fixed here. I updated this PR to close #3077, kept #3005 open for its separate idle-qubit detection bug, and renamed the release-note fragment accordingly. Thanks for flagging this.