Better file handling

[rvosa]

The pipeline operates on a SQLite database file that is progressively populated by a chain of rules. Because SnakeMake gets confused about time stamps and rule dependencies when the same file is modified by multiple rules, separate (0-byte) files are created. This is messy. Supposedly, the SnakeMake team is working on facilitating the SQLite use case in upcoming releases. It would be good if the pipeline could be made cleaner in that way. In addition, the handling of other files should also be improved. Users paint themselves into corners with poorly configured run attempts and then don't know what to remove to try again. This should be more fool-proof, possibly simply by adding a clean rule. Even better would be if not all intermediate results were lost, although that's harder to do correctly given certain failures.