fix visiblity determination for subtomo particle reconstruction

[daniel-ji]

When doing subtomogram reconstruction, visibility of the extracted particles that are to be backprojected is inconsistently / incorrectly determined based on the box size after binning and does not account for the fact that the actual extracted box from the tiltseries is larger than the output box size when a bin factor is set.

The visibility should be based on the extracted dimensions before binning (so bin factor * box size, i.e. binning * s), to align with the later TomoExtraction::extractAt3D_Fourier call that extracts at the s02D dimension (just a couple lines after).

This also then aligns with the logic done in subtomogram extraction as well (using the pre-binned dimensions to determine particle visibility): https://github.com/3dem/relion/blob/master/src/jaz/tomography/programs/subtomo.cpp#L733

Note: I'm still working through a bit of debugging as for why my re-implementation (done to support the reading of ZARR files & files on the cloud in a chunked manner) does not match numerically with RELION's output, so there might be more small bugfixes I suggest in the process of trying to align the two implementations (this code is will soon be open source). So feel free to hold off on a potential merge if you would rather not have me continue to open one-line PR changes, sorry about this. I thought I'd at least open this PR first, to bring attention to this small inconsistency. Thanks so much!

[daniel-ji]

Due to my classes starting back up again and other work, my debugging has been paused, so feel free to merge this in if it seems fit, thanks!