Should we just rely on the ending to determine the filetype?

[krystophny]

Should we just rely on the ending to determine the filetype?
On Linux the ending can be arbitrary (even missing). Do we force the users to use the ending? What if data.in is actually a HDF5? For text based files there are a variety of sensible endings (.txt, .csv, .tsv)...

1. Using python-magic or calling file via the OS we could determine the input file type from the file header. (only on Linux though I think)

2. We could just try all filetypes until the first success.

3. We could add a config option for the filetype (also a good idea in addition to 1 or 2)

For saving only option 3 would be possible (unless the file already exists, but we don't want to overwrite it anyways?)

Originally posted by @Rykath in #153 (comment)