Dmk

- Added `dummy` cli command
- The number of randomly added / removed dummy blocks is now proportional to
the vault size

- Replaced KDF to Argon2 (was Scrypt)
- To identify the block, we now try to decrypt and verify its header.
Previously, two hashes were used for this. It was more complicated,
less efficient and less reliable
- Random padding is no longer added to the beginning of the block
- Fixed `setup.py` issue: subpackages were not installing

- Replaced random sized blobs with fixed sized blocks
- The vault file is now indistinguishable from random data

- A binary "vault" file is used instead of a directory. This is more reliable

- Entries and fakes are stored as blobs in the vault file

- Blob sizes are random. The size of the fakes is random. Large entries are
broken into parts, small entries are padded

- (flaw) The file stores blob sizes openly. It can be easily recognized

- All imprints have been moved inside the files

- The file names are now random

- Elements are split into a random number of parts

- Directory now stores random salt

- The directory is indistinguishable from random data

- The data is stored in a directory. Imprints are encoded in file names

- Random padding is added to real data files

- Random fake files are generated to hide the number of real ones

- (flaw) It is easy to guess that the directory was generated by the utility:
filenames have a fixed length