Logger 0.2

Need to:

1. design meta data file in YAML. This file will contain the information about the scheme, parameters, ... used
2. discuss about header (amount of bytes vs data type). In the dump file, we should write some information in the header. Currently we write the number of bytes for each mask but maybe it would be better to use an enum.
3. MPI: Should write 1 file per rank in order to keep a nice parallelization. Still need to figure out how to do it.
4. Flipping offset on the fly. We currently reverse the offset afterwards, could be faster to do it on the fly.
5. header broken? For the offset, we write only 7 bytes and depending on the endian, it may be broken. Same for the reader. need to fix todo in logger.c (bytes -> bits)
6. API. We need a good interface to specify the writing of the data. It should be easy on the user side and the SPH/Grav/... developer side.
7. restart. When restarting, we need to go back to the correct timestamp and start again.
8. stop writing consts every step. This information is useful only at the beginning, therefore no need to write it down later.
9. User defined frequency of writing for each field. In order to make our life more easy, we can ask the user to provide only multiples of the highest one.
10. move reader to pyswift or swift
11. do stars and gparts
12. write index files in binary. We need to move away from HDF5.
13. figure out the names of the different object
14. part_is_starting. We need to deeply think about the timestep at the writing, we may need to use part_is_active
15. add logger and snapshot policy. We need a way to let the user decide between snapshot and index files.
16. decide if snapshot are allowed + different variables for index. The variables of the index should not be the ones for the snapshots.
17. do not drift particles when writing index. As we do everything exactly as in the snapshot, we drift the particle before an index file which is useless.
18. Write the full time and not the integer
19. Logger should be an independent library
20. Add debugging flag at beginning of each data chunk
21. Remove logger_parameters
22. Need to find a way to deal with cosmology (time_base works differently)