[mceachen]

I had this same issue.

PhotoStructure can read iPhoto libraries (and LR XMPs), and will coalesce both "sidecar" dupes (like JPG/RAW pairs), as well as images that were downsampled from other photo services (like Google Photos). It can leave your originals where they are, or copy them into a single time-stamped folder hierarchy (yes, you can edit the format of the timestamp if you don't like the LR standard). Files with the same SHA are not copied, but referenced in your library db.

Your library's metadata is kept both on disk (in XMP sidecars) as well as a sqlite db for fast access, and you can run queries on the db to do anything more exotic.