Tutorial on structure correlation and generation

Description

In the past we have repeatedly discussed structure generation schemes beyond enumeration. There is an example (test_cluster_vector_correlation.py), which takes a few steps in this direction but is not useful for the average user. With 8f7ee241 I removed this example along with a few other examples from the user guide (although both scripts and documentation text are still present in the repo).

The objective of this issue is to provide a (more) comprehensive analysis of structure generation and to include a discussion of this aspect in the user guide. To this end, I suggest the following approach

• develop a protocol for the generation of structures with specified correlations (#257 (closed))
• write a section that illustrates the use of this feature
• add a section under advanced topics on structure generation, which integrates the existing structure enumeration section with the section that describes the SQS functionality

The term SQS is usually meant to refer to special quasi-random structures, which implies minimal correlation. (In the case of a binary alloy at 50% this means all elements of the cluster vector should be zero except for the first, which is always one). One can, however, easily generalize the principle to any other correlation, which could be utilized to generate a well balanced set of input structures for systems with large unit cells. One can anticipate several versions of such an algorithm, which should be tested and refined.