Insilicoseq

This is a minor bugfix release:

* Fix python2 compatibility error for downloading genomes from the ncbi ( 45 )
* Fix python2 error handling when a genome fails to download

* Fix crash for when a genome is shorter than the template length ( 45 ):
now uses a simpler error model and allow negative insert size. Will crash (rightfully) is a genome is shorter than the read length

This is the first stable release if InSilicoSeq 🎉

Changes:
* `--model` is now case insensitive
* `--kingdom`is no longer limited to a single kingdom and now accepts a combination of bacteria, viruses and archaea

Bugfixes:
* fixes a lot of error handling that were python3 specific and made the python2 version crash

Bugfixes:

- `iss generate` now saves the abundance in a file ( 36 )
- `iss generate` now prints a warning if a user try to use an error model in `--mode basic` ( 38 )
- fixes a bug that crashed the app with the default number of reads ( 40 )
- fixes a bug that sometimes prevented the software to cleanup properly ( 41 )

Thanks to jhayer for a help! 🚀

This release brings bug fixes and small enhancements.

Changes:
* deprecation of the 1D KDE. 2 Dimensional KDE is now the only other error model available together with the basic one
* the option `--model` has been renamed `mode` and is KDE by default (see 30 )
* the option `--model_file`has been renamed `model` and uses the HiSeq error model by default
* the `--n_reads` option now allows suffixes 'G', 'M' and 'K' both in lower and uppercase. (see 29 )
* `iss --version` and `iss -v` now print the version number and exit. (see 27 )

Bugfixes:
* the temporary files are now cleanup up if the user interrupts `iss generate` ( see 31 )
* the empty genomes are now correctly skipped if you use the `--ncbi` option (see 32 )
* the `--gc_bias` options is functioning again ( see 33 )
* the `NovaSeq` and `HiSeq` error models are now really usable (see 34 )

This releases mainly adds error models.

Changes:
* added a NovaSeq error model
* updated MiSeq and HiSeq error models
* counting bam records is now way faster
*

Bug fixes:
* fixed a bug that affected reads mapped against the reverse complement
* changed the matrix transposition method leading to smaller file size for the error models

This release includes changes to the `iss model` module.

Changes:
* `iss model` now constructs its statistical model on 1 million reads instead of the entire bam file
* no loss of base quality accuracy could be detected
* greatly speeds up model generation (~20 minutes against ~100 minutes before)
* reduces memory usage