During the execution of CheckMetadataConsistency we currently load the entire chunk list for a specific collection in memory in order to check them for consistency.

For collections with millions of chunks, this means that CheckMetadataConsistency would possibly need GBs of memory to load all the chunks in memory.

Instead of loading all chunks at once in memory, we can use a cursor and fetch small batch each time. Like we do for mergeAllChunksOnShard.