I believe this has been already talked in (https://forum.inaturalist.org/t/import-csv-photo-link/68377)
You will probably need to create a script in RStudio, Python, or another programming environment to associate the observations in your CSV file with the correct image files, potentially through the iNaturalist API or standardized file naming.
The main challenge would probably be linking each photo to the correct entry in your database. Dates alone may not be enough, especially if you have multiple specimens photographed on the same day.
The easiest approach is usually to use a consistent naming system for the images that matches a unique ID in your CSV file (for example: observation IDs, trap/sample IDs, etc.). Then you could automate the matching process with a script in R or Python.
For datasets as large as yours, automating the workflow is definitely worth it.