fix_length_mode="trim" can lead to confusing loss/metrics

When using `fix_length_mode="trim"` in DCUNet, the signal is right-trimmed to the next possible size, and then zero-padded on the right to reconstruct the original size. When using a loss or metric that does not know about the actual input size to the model, the loss/metric values can be way off, since essentially a chunk of the expected signal is zeroed in the estimate. (I don't think it hurts training though.)

What to do? Maybe just add note to the docs?

The issue names DCUNet and fix_length_mode="trim" but no files or tests. Start by locating the DCUNet documentation and the trim behavior, then verify how losses and metrics receive the zero-padded estimate; done should be a clear documentation note explaining the input-size caveat and its effect on evaluation.

Written by the indexing model from the issue text.