coverage: 40.46% (+0.002%) from 40.458%
when pulling c428014 on python-envs
into 8b9d113 on master.

With this PR is possible to run a different version than 3.8 (the jep shipped with grobid requires python 3.8) and relay on conda for the environment.

Here what I did for running DL models natively using a conda environment:

1. cd grobid
2. Switch to this branch
3. rm -rf grobid-home/lib/lin-64/jep
4. cd ..
5. git clone https://github.com/kermitt2/delft (delft should be in the parent directory)
6. assuming conda is installed (if not, I installed this) and the environment delft has been created with either python=3.8 or python=3.9 (e.g. conda create --name delft python=3.8
7. cd grobid
8. pip install delft==0.3.3 (with deflt==0.3.4 you also need to install blingfire pip install blingfire
9. conda install jep=4.1.1 (conda will install also openjdk 8)
10. conda uninstall --force openjdk-8 - or conda uninstall --force openjdk (if you use miniconda3)
11. conda install --force openjdk-17 or conda install --force openjdk=17 (if you use miniconda3)
12. change any model in the grobid config to use delft instead of wapiti (e.g. header model)
13. ./gradlew run