Fix misalignment error by padding variable-sized frames in VQT computation

[juliohsu]

Added padding logic to ensure consistent frame sizes across octaves when the hop_length is not a power of 2. This prevents size alignment errors by padding smaller frames to match the maximum time axis frame length before concatenation.

[gudgud96]

Hey @juliohsu, just saw this. I helped with the VQT part some time ago.

Thanks for spotting the edge cases! Indeed there might be rounding issues if hop length is not a power of 2. I think there might be a way to make get_cqt_complex more robust towards this case (I am not sure), but I think your solution using interpolation should work too! LGTM.

[juliohsu]

@gudgud96 do you think there is any thing to be test or improve, in order to be merged?

[gudgud96]

@juliohsu I'll suggest to (i) remove unnecessary comments; (ii) add a test case under test_vqt.py that would only pass, if the interpolation line you proposed is included.

cc @KinWaiCheuk