Bind audio pre-roll claims to the bytes that were materialized

Declaring an audio pre-roll duration in training configuration is insufficient. The crop materializer must physically include the prefix, record requested and effective pre-roll separately, and bind those values into the crop receipt. Downstream training and inference should recompute effective pre-roll from alignment-start minus audio-start and reject mismatches. Source-boundary clipping is valid when explicitly represented; silently treating total crop duration as pre-roll is not. This verifies materialization and policy identity, but does not prove the added context improves recognition.

Applies when: matched audio preprocessing is required across training, development, and inference. Keywords: audio pre-roll crop bounds materialization receipt training inference ASR