Following #2415, two points that were not tackled inside #2476 in the interest of short cycles

• Add a section in 2.1 on how to determine chance with a Decoder API (waiting for #2147)
• Rethink 2.2 section on OVO vs OVA, not sure it is useful as Decoder doesn't provide this flexibility easily. But show how to plot a confusion matrix using Decoder on a simple example.

The date and time on the output of the running docker container isn't correct, adding a volume "-v /etc/localtime:/etc/localtime:ro" might fix this?

This would help me know how long a process

Some is already in .gitignore, but some is not. We should check all examples.

This line https://github.com/SIMEXP/niak/blob/master/bricks/fmri_preprocess/niak_brick_regress_confounds.m#L234

should say "Adding motion parameters..."

and not "Adding high frequency noise..."