allow users to continue model version creation later

[LiFaytheGoblin]

will fix #31

• adds a button to continue a model version manually or automatically
• enable continuation of a model later by attempting to lazy load evidence from the db if it is not in the cache. If it's not in the db either, the evidence is newly created.
• adds tests

[LiFaytheGoblin]

Next steps:

• [x] show button only for unfinished versions where no adhoc task is still running
• [x] skip dataset gathering
  • [x] implement method to restore dataset from file system
• [x] skip training test data split
  • [x] implement method to restore trainining and testing datal from file system (same as dataset?)
• [x] skip training
  • [x] implement method to restore trained model from file system
• [x] skip predicting (?) -> then the model is already finished! throw a logic error!
• [x] catch case where we try to get related data but have lost the idmap.

New tests:

• [x] model version creation: do steps 1 to x manually, then call create() and make sure the version creation is finished successfully.
• [ ] test restore_raw_data for dataset
• [ ] test restore_raw_data for training data
• [ ] test restore_raw_data for test data
• [ ] test restore_raw_data for model

[LiFaytheGoblin]

Next steps:

• [x] make sure continue version button works for manual creation as well.