Neuroconda 2.0 - shell wrappers, more sustainable development model

• New development model where we separate key package requirements (neuroconda_basepackages.yml) from the final neuroconda.yml, which also includes any dependencies of these packages. This makes maintaining and updating a lot easier and hopefully more sustainable (e.g., packages that are no longer a dependency for any base package will get pruned in the next update).
• Automated update and install with GNU make. This was a lot of work, but it should make it easier to push out minor releases frequently in the future.
• Introduce CBU-specific shell wrappers for adding non-conda packages to path (these can be used as templates for your local setup). This is necessary because the env_vars.{sh,csh} stuff in conda remains quite broken (see README for more on this). Maybe one day it will get fixed and we can get back to a conda-only setup.
• Thanks to above improvements in the development model, neuroconda packages are now bleeding edge. We include the latest available release of Altair, Nibabel, Nipype, Nilearn, etc etc... The only version left behind is python, which is still 3.7.3.