Changing the project configuration deletes all activities

[esyriani]

Describe the bug
ReLiS is supposed to support iterative systematic reviews. Therefore, we should be able to change the projection configuration (for example the classification scheme). It should allow me to add new categories, delete exiting categories, or update the configuration of existing categories. However, currently, if I re-generate my configuration and re-deploy it, it erases all previous activities: screening, QA, and classification data. Only the papers, authors and venues remain.

To Reproduce
Steps to reproduce the behavior:

1. Create an initial configuration
2. Screen one paper
3. Classify that paper
4. Modify the configuration
5. Go back to see the screened papers and classified papers, or any statistics about screening and classification. Everything is erased

Expected behavior
If we want to "reset" the project, then we can delete all papers (see #19). But modifying the configuration should not delete the previous information. It should try to modify what exists. Especially for the classification scheme. We should allow anything the database allows to modify. So if we are modifying a table that has rows, we let MySQL tell us it is allowed or not.

[esyriani]

Add the keyword keep_screening right after SCREENING in the configuration file to only modify the classification part, and keep the screening part untouched.