Description

When adding an image, it would be useful to detect if the image is a duplicate and instead of adding the image again, return a reference to the existing media.

We have a header that gets added in some large workbooks to each sheet, and this substantially increases the file size. If we open the file in Excel and save it again, Excel detects the duplicate images and consolidates them on save, reducing the file size.

I think it would be possible to change addMedia in picture.go to handle this behavior. AddPictureFromBytes would change slightly as well:

1. addMedia steps through all the saved media and looks for byte slices that have the same length as the file we are trying to save. Compute the hash for both (we can obviously re-use the hash of the file being saved for future comparisons)
2. If the hash matches, return the media path for the existing image. If no matches are found, save the new media and return its path
3. AddPictureFromBytes calls addMedia before calling addDrawingRelationships and uses the media path provided by addMedia in the call to addDrawingRelationships.

Since checking the length of a slice is a constant time operation, and very few slices should have the exact same length without actually being the same media there shouldn't be much of a performance impact from the hashing. However, it would be worth adding a benchmark to ensure this doesn't cause a regression for #274. It would also be useful to incorporate a benchmark for actually saving the xlsx file since it's likely that the performance impact of this check would be offset by not having to write as many files.

I will wait for feedback on this one since while I think it may be useful but I understand not wanting to risk the performance impact.