chore(bb): refresh pinned Chonk IVC inputs (fixes nightly debug build)

[AztecBot]

What

Refreshes the pinned Chonk IVC inputs to 49442f24d1be75a0 (barretenberg/cpp/scripts/chonk-inputs.hash). The new tarball was regenerated and uploaded by the ci-refresh-chonk post-action from the current protocol circuits, and self-verified by proving the smallest refreshed flow in C++ and bb.js (which re-runs the exact failing test). Single commit, hash-only change.

Why — Nightly Debug Build failure

Nightly Debug Build failed every night for ≥4 days (runs #109–#112, Jun 6–9), latest on next HEAD d9b6e71:

• Run: https://github.com/AztecProtocol/aztec-packages/actions/runs/27188115040
• Failing test: bbapi_tests ChonkPinnedIvcInputsTest.AllPinnedFlows

Root cause — stale pinned inputs

Accumulating circuit 10/10 (hiding_kernel), the old pinned VK (209dde8e69a27c9f) no longer matched the circuit built from current code:

Pub inputs offset mismatch: 669 vs 673
Commitment mismatch: Q_M / Q_C / Q_L / Q_R / Q_O / Q_4 / Q_5 ...

The kernel public-input layout grew by 4, so KernelIO::reconstruct_from_public read misaligned fields and the databus consistency check in chonk.cpp (process_public_inputs_and_consistency_checks) tripped:

Assertion failed: (kernel_return_data_match)
Reason: kernel_return_data mismatch: proof contains
  { 0x1832...e7ea4, 0x0 } but kernel_calldata commitment is { 0x..01, 0x..02 }

{0x..01, 0x..02} is the BN254 generator (an empty databus column) vs real return data in the proof — the "pinned inputs drifted from the protocol circuits" signal. It surfaces only in the debug nightly because the check is BB_ASSERT_DEBUG (compiled out of release; chonk_inputs.sh check runs --disable_asserts), which is why merge-queue CI stayed green while breaking protocol-circuit PRs merged without ci-refresh-chonk. No bb product code is changed here.

Notes

• Carries --ci-skip (hash-only refresh, already self-verified by the refresh post-action), matching the standard refresh-PR convention.
• The earlier prior attempt chore(bb): refresh pinned Chonk IVC inputs after protocol-circuit changes #23744 never ran because it was left as a draft (ci3.yml gates the refresh job on draft == false || ci-draft); this PR used ci-draft so the refresh executed.

Ready to merge to next once marked ready-for-review.

[AztecBot]

Pinned Chonk inputs refreshed to 49442f24d1be75a0. The update commit includes a head-commit --ci-skip marker so the automatic follow-up run does not repeat CI.

[AztecBot]

Automatically closing this stale claudebox draft PR (no updates for 5+ days). Re-open if still needed.