Measured audio-edit findings: no non-speech removals found; correct the earlier speech claim; add scan manifest

Add 6 audio-visual removal samples (voice stripped), speech-band metric, two-regime explainer

Add 9 more add_and_remove samples (10 total) + at-a-glance edit-strength table

InsAVE-80K sample viewer: 5 eval + 5 train input/output pairs

initial commit