Funannotate

- update to the way that `funannotate predict` parses `maker2` results, now using maker models directly as opposed to pulling out annotation from each predictor.
- bug fix if running `funannotate compare` with a single species

- fix to the `braker1` method where augustus output was not properly found
- minor update to `--optimize_augustus` training to align with method used in `braker1`

- fix issue with parallel `augustus` where very large scaffolds would cause large memory usage, script now chunks the data into 500 kb sections with 10 kb overlaps on each side, runs in parallel, and then combines the results.
- re-ordered transcript evidence in `funannotate predict` to address providing hints to `augustus`
- some minor bug fixes

- build a check for `augustus` version and test if it will function with `busco` and `braker1`
- revamped `busco` mediated training of augustus to run `busco` quickly, filter evidence data corresponding to busco models, filter genemark-ES data, run `evidence modeler` to get high quality gene sets, filter EVM output with busco to build a final augustus training dataset, and finally train `augustus`
- improved system info reporting
- due to problems with installing `augustus` on different operating systems, `augustus` is not installed default via `brew install funannotate`. However, running `funannotate predict` without a version of `augustus` installed will give you some hints on how to install it for your system.

- important bug fix for `augustus`, previous versions were running `augustus` with the stop codon inside the prediction, which results in the gene models to fail validation in `evidence modeler`, thus this update is recommended for all users
- added high quality `augustus` models to be pulled out of annotation if they are represented by evidence using the `--hintsfile`, these models are passed to EVM with additional weight
- fixed `genemark` bug where a single contig resulted in an error, thus `funanntoate predict` can now handle a single contig as input correctly.

- improvements to the gene model filtering in `funannotate predict`, ability to keep gene models without proper stop codons if desired, `--keep_no_stops`
- `augustus` is now multi-threaded
- upgrade of packaged `BUSCO` to v1.2, slightly faster runtime and simplified code
- non-fungal options are now included in `funannotate`, however use with non fungal genomes has not been extensively tested. Options of note are `--organism`, `--busco_db`, `--eggnog_db`
- secondary metabolism enzymes added to `funannotate compare` if genomes were annotated with `antiSMASH`