Calculate image statistics for basic QC

**Is your feature request related to a problem? Please describe.**
Since all of the data is read in this code base as part of the OME-Zarr conversion, it is a good opportunity to calculate and store some useful image statistics, such as:
- Numbers/percent of saturated pixels
- might be most useful to do it on the "1" pyramid level
- A few useful quantiles for image brightness setting in visualizers
- Actually, just a histogram of each channel with bin_width=1 instead of first two bullet points (?)
- mean of first image in first tile of each channel (dark field sanity check)

**Describe the solution you'd like**
Alter the PNGReader class to do this as it reads the delayed images?

**Describe alternatives you've considered**
Doing this elsewhere requires reading all images. On the HPC, you can do it at about 70 ms / image / core, but orchestrating another SLURM job before data upload is complicated.

Start by locating the PNGReader class and its delayed-image read path. Before implementation, resolve whether to store saturation and quantile statistics, per-channel unit-width histograms, the first-tile dark-field mean, or a defined subset, and where the statistics belong in the OME-Zarr output. Done means the agreed statistics are calculated during conversion without a separate full-image read and are present in the converted output.

Escrita pelo modelo de indexação a partir do texto da issue.