Add docker and solution to vectorfield error

[gatocor]

This pull provides solutions to three aspets:

• Make a requirements.yml in the main folder so everything can be installed in a more straightforward way with conda.
• Make a docker folder with registriationtools, vt-python, simpleitk and skimage uinstalled so one can work out of the box with registration packages without going over the pain of installing it.
• Make htat if the klb package is not installed, save the vectorfield in nonlinear registration be stored in tif format instead. This avoids the problem with this outdated format that I have not been able to install anyhow.

[leoguignard]

Small question about the number of dependencies. Otherwise all good. Thanks a lot!!

[leoguignard]

Not that it's necessary wrong but it looks like a very long list of dependencies like for example is networkx really necessary?
But anyway, since it's already built it does not matter too much.

[gatocor]

This is just because skimage is installed.

I added some aditional basic packages to help visualizing the results and working around with the data (matplotlib, scikitimage), an those installed all the rest (networkx for example is a installed by scikitimage)

With the docker is the same, I even installed SimpleITK to rich set os tools without having to install anything else.

Do you think a more minimalistec version with just registrationtools will be better?

[leoguignard]

Hi, yes I think that if there is no direct use of a library it's better to leave it out.
That being said, I am not super familiar with docker builds and I might be wrong

[leoguignard]

Same comment as for the other dependencies file