**issue:** Handle `nan` correlation values when using `np.corrcoef` for belt similarity

[Frix-x]

issue: Handle nan correlation values when using np.corrcoef for belt similarity

np.corrcoef still returns nan when either PSD is constant/degenerate. That propagates through to similarity_factor and the printed similarity (nan%). Please handle this explicitly (e.g., check np.isnan(correlation) and map it to 0% or a clearer message) before converting to a percentage.

Originally posted by @sourcery-ai[bot] in #265 (comment)