Ceph » bluestore

Originally 1M was chosen as the bluefs_alloc_size since metadata is stored in rocksdb, which persists it in large chunks. In more recent testing, we found no performance impact from using bluefs_alloc_size 256k. Reducing this further, so that both BlueFS and BlueStore allocators use the same allocation size, will enable BlueFS to use any free space left, so fragmentation will no longer lead to an out of space error unless the device is actually fully used.