Duplicate uploads might happen with AbstractSharedCachingDataStore

If a file is staged for async upload in UploadStagingCache and then another call to AbstractSharedCachingDataStore.addRecord is made for a file with same SHA1, the new call goes directly to the backed to write the file, because the cache is not taking into account pending uploads. This makes 2 uploads to happen for the same blob: one async (from UploadStagingCache) and one sync (from AbstractSharedCachingDataStore.addRecord

(cc amitjain)