p.s. just be aware of conflicts with @adamemerson's work in #47422 and #47566

This pull request can no longer be automatically merged: a rebase is needed and changes have to be manually resolved

this is a good start! we found all the places that need version tracking

as a next step, we need to arrange for each datalog shard to have its own instance of RGWObjVersionTracker, so it can use that same instance for every step of the process (Read, Init, List, and Sync). this way, we can detect racing writes at any point of the process, instead of just in between the new reads and their writes

RGWDataSyncCR has a member variable rgw_data_sync_status sync_status that stores a map of the per-shard sync markers, which gets passed to each of these coroutines RGWReadDataSyncStatusMarkersCR, RGWInitDataSyncStatusCoroutine, and RGWListBucketIndexesCR. each RGWDataSyncShardCR also gets a pointer to its own shard marker

we'll need to do something similar for the storage of these RGWObjVersionTrackers. for example, by adding a std::vector<RGWObjVersionTracker> to RGWDataSyncCR that gets passed along with the sync status and shard markers. RGWDataSyncShardCR should get its RGWObjVersionTracker& by reference, and share that with its RGWDataSyncShardMarkerTrack

Addressing this in a rebased version #47682

Closing against #47682