funannotate v1.7.0

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