Data type reading in .npy files

[corentin-prigent]

This update allows hex2tet to read the data size contained in the file, in order to read data properly.

[Algiane]

Thanks for this fix.

• I think that we need to improve the readibility before merging this : I propose to define macros at file begining to allow the user to easily understand the meaning of the hex code and to add at least comments on what is readed in the header.

• Is it possible to add 2 minimal .npy files with different data sizes to illustrate this PR (so we can easily see that the develop branch fails to read a file with a wrong data size).

[corentin-prigent]

• Several macros have been defined and some comments have been added in order to increase the readability of the reading process for .npy files.
• Reading process for data size has been modified in order to make it more portable and more general. We now look for the ket "descr" in header dictionnary to locate data size specifications.
• Several files are linked with this message: they illustrate the need for proper data specification. example-i2.npy and example-u4.npy are two .npy files containing a 4x4x4 array of numbers ranging from 1 to 64, one encoded in 2-byte signed integers and the other in 4-byte unsigned integers. example-i2-devel.mesh and example-u4-devel.mesh are the outputs with the current develop branch. In example-i2-devel.mesh, it can be seen that reference values are wrong, while they are correct in example-u4-devel.mesh. example-i2-fix.mesh and example-u4-fix.mesh are the output files with the proposed fix. The two files are identical and correct.
  example.tar.gz

[Algiane]

Perfect, thanks!