Include an internal check for empty prokka fasta files

[papanikos]

Hello @deyvidamgarten ,

I run into an issue where, for some of the contigs I was analyzing with marvel, prokka was producing empty protein fastas. These caused later hmmer searches to fail and the whole marvel workflow was failing.

This pull request addresses this issue.

• I included a small cleanup step of the intermediate prokka output so that the analysis keeps going. If everything is removed exit and warn the user.
• I included an environment.yml that freezes some dependencies to versions that worked for me. Since this was published, packages have been updated and the latest versions of biopython and scikit-learn throw errors. This is meant to be used with conda.
• A brief description of how to use the environment.yml with conda is also included in the README.

I tested it with some of my broken files and it works as expected, at least the erroneous bins are skipped.

[deyvidamgarten]

Hello Nikos,

Thank you for addressing this issue. I really appreciate it.

Cheers