the mnist-convnet.py examples was overloaded. This commit splits the
file into the most minimal mnist-convnet and mnist-tfslim. Further
it adds an example about the visualization of filters which
illustrates the usage of `VariableHolder`.

In addition, this also adds a boilerplate template for projects,
which might facilitate the process of starting a new project.

After writing `x = Conv2D(...)` one can access the used variables
by `x.variables.W` and `x.variables.b` if they exists.

This is much better than writing:
`x.name.replace('/output:0', "/W:0").replace('tower0/', '')`

whould would also break the InferenceRunner.