Code has been repackaged to conform to a "proper" python package -- which means it now also resides on PyPi and a Bioconda package can be built. Along with the repackaging there are many improvements/fixes.
* funannotate now keeps track of "trained species" for all of the ab-initio gene predictors (Augustus, gene mark (optional), snap, GlimmerHMM, codingquarry). This requires all users to update their database, ie `funannotate setup` command. After running `funannotate predict` the software will output a JSON file containing the paths to the trained parameter files -- this can be used again for a different genome via the `funannotate predict --parameters` options. This parameter file can also be added to the database with the `funannotate species -s genus_species -a parameters.json` command. Running the `funannotate species` command will output a table in the command line of which species have training data. Addressed 320
* antiSMASH remote script fixed and parser updated for v5 output.
* added filtering for gene models that start/end in a gap that can sometimes show up after running `funannotate update`
* added a check for `diamond` version of the database and current copy -- this results in many hidden errors by users, ie diamond databases were created with an older/incompatible version than what is running currently.
* updated `Augustus` functional check
* removed `RepeatModeler/RepeatMasker` as strict dependencies. Due to RepBase change in usage license, repeatmasker/modeler are not available to most users. The `funannotate mask` command can still run this routine if you have the necessary dependencies installed, however, the current default is simply to run `tantan` masking. This is probably not sufficient for most genomes, thus happy to integrate a robust solution once one exists for repeat masking.
* augustus parameter training now done in the local output folder, so no longer need write access to $AUGUSTUS_CONFIG_PATH