cmake meta example organisation

[karlnapf]

Currently, we run a script that translates all meta examples (single cmake command), outputting something as

Generating examples from meta-language
Translating tests/matrix_types.sg
Translating tests/vector_types.sg
Translating gaussian_processes/gaussian_process_regression.sg
Translating gaussian_processes/gaussian_process_classifier.sg
Translating binary_classifier/lda.sg
Translating binary_classifier/linear_svm.sg
Translating binary_classifier/kernel_svm.sg
Translating regression/kernel_ridge_regression.sg
Translating regression/linear_ridge_regression.sg
Translating regression/multiple_kernel_learning.sg
Translating regression/random_forest_regression.sg
Translating regression/support_vector_regression.sg
Translating clustering/kmeans.sg
Translating clustering/hierarchical.sg
Translating clustering/gmm.sg
Translating multiclass_classifier/large_margin_nearest_neighbours.sg
Translating multiclass_classifier/quadratic_discriminant_analysis.sg
Translating multiclass_classifier/random_forest.sg
Translating multiclass_classifier/multiclass_logisticregression.sg
Translating multiclass_classifier/multiclass_linearmachine.sg
Translating multiclass_classifier/svm.sg
Translating multiclass_classifier/shareboost.sg
Translating multiclass_classifier/relaxed_tree.sg
Translating multiclass_classifier/gaussian_naive_bayes.sg
Translating multiclass_classifier/knn.sg
Translating multiclass_classifier/linear_discriminant_analysis.sg
Translating multiclass_classifier/multiclass_ecoc_random.sg
Translating neural_nets/feedforward_net_classification.sg
Translating neural_nets/feedforward_net_regression.sg

Instead, it would be better to have every of those translated examples as a single cmake command (which is parallelised, handled by the progress bar, etc etc)

Easy cmake refactoring.

@sorig

[karlnapf]

This will also allow to enable the '-DTRAVIS_DISABLE_META_CPP' flag on the modular travis builds to save time

[karlnapf]

Solved