update execution id

[tdiprima]

Currently, when running the pipeline, you need to make sure that you're not running the new training prediction phase with the same execution id as the previous one. Such that whenever we update our CNN model, we need a new execution id assigned.

We probably shouldn't leave this up to human memory.

It would be good if at some point we ping the database to check to see if we already have data for the version specified in variables.sh. And alert you if it's the same, so you can change it.