Expelliarmus

In this release we have some new features:
- it is possible to read a file in windows expressed in milliseconds, using the new `read_time_window()` generator.
- it is possible to save a NumPy array to DAT, EVT2 and EVT3 formats. EVT3 saving does not support events vectorization yet, but this is usually not an issue in terms of file size, since few recordings take advantage of vectorization in general.
- new benchmarking data!

In this release:
- `Muggle` has been merged with `Wizard`.
- C code has been _hugely_ improved, providing a substantial performance bump and more robustness.

Now two classes are available: `Wizard` allows to read files entirely and to cut them to a desired duration; `Muggle` allows to read a file in chunks when it is too big to fit in memory.

Now, thanks to `Muggle`, you can read a large file in chunks! Since a simple *muggle* cannot use magic, even less a powerful spell such as *expelliarmus*, he/she needs to read the file a little bit at time!
`Muggle` employs a generator method that allows to read `nevents_per_chunk` events at time from the file.

Now, it is possible also to easily change the order of the coordinates: by specifying a different _dtype_, one can change the order of timestamps, polarity and coordinates. This feature has still to be refined, though, bu the C code is perfectly capable of doing that.

In this release:
- `read_FILETYPE` functions to read DAT, EVT2 and EVT3 to structured NumPy arrays.
- multiplatform testing on Ubuntu, MacOS and Windows.
- `cut_FILETYPE` functions that lets you choose the minimum recording duration to which you want to limit a certain recording (see README).
- CI pipeline set thanks to biphasic.