Smudgeplot

- fixed logging (now it's directed to err stream)
- an estimate of ploidy based on all smudges of that ploidy (instead of the ploidy of the brightest smudge)
- smudgeplot interface uses `.py` suffix to meet community standards

This version is using the same computational backend as the previous version (0.1.3), but it's wrapped in a single interface that is expected to be kept in future:


smudgeplot <task> <arguments>


Further adjustments:
- improved algorithm for placing smudges on the plot for higher ploidy levels than 4
- alternative algorithm for extracting kmers available (`--middle` in `hetkmers` task)

I had no idea how to name the release, so I have decided to name individual versions of smudgeplots by [types of windows](https://modernize.com/windows/types), so let's start simple: Single-Hung it is. Hopefully, it will be good enough name to carry all the smudges.

- no quantile filtering by default, a new parameter `-q` to set up the filter, the falg `--no-qunatile-filtering` was removed (set `-q 0.99` to set up the previously default filter)
- algorithm for peak identification was failing if AAB was absent but AAAABB was present. This problem should be resolved by now by considering both the diploid and triploid 1n estimates.
- fixed the installation instructions
- minor fixes

- switch to colorblind friendly palette
- parameter nbins is autoscaling by default and fixed if defined by user
- added an option to disable quantile filtering (`--no_quantile_filt` flag)
- improved interface & README

smudgeplots are right now

- production ready (fingers crossed)
- scaling for any genome size and read set (tested on a dataset of 1,000,000,000 kmers)
- fully automated (tested on set of 25 obscure genomes)
- equipped with rich set of warnings whenever the inference seems strange

This is the first draft of automated smudgeplots - plots that reveal genome structure without any assumptions about assembly, mapping and base quality scores.

Right now we have:
- non-optimized python script to extract kmer pairs
- Rpackage and R scirpt to generate a smudgeplot out of the extracted kmers
- a small R smiplet to guess L/U thresholds for kmer filtering

Everything should be operational, but it was not extensively tested