Gadma

- Fix bug with momentsLD engine for models with exponential or linear change
- Restrict momi2 to use only one core
- Use another scipy function for Nelder_mead algorithm (local)
- Make all histories with different parameters when structure is changed (e.g. from 1,1 to 2,1)
- Allow usage of models specified using gadma API

* Fix bug when failed model is reported as best (Issue 73 )
* Add additional checks for Bayesian optimization
* Add documentation about BO

- Add linear extrapolation for dadi (Issue 68 ).
- Engine momi is renamed in momi2.
- Add dominance rate in the params file interface.
- Fix bug with CLAIC evaluation (for any eps it was taken as 1e-2).
- We changed hyperparameters of GADMA2
- Fix minor bugs.
- Fix documentation

This is rather small release with some fixes.

- We have removed bug that was caused by new moments version.
- GA now can be run after BO
- Momi engine can read .gz files now.

1. We present a new engine - ``momentsLD``! It is the first engine that use LD statistics for demographic inference in GADMA. For more information please see [documentation](https://gadma.readthedocs.io/en/latest/user_manual/set_engine.html). As in case of the ``momi`` engine if there are some problems with usage of momentsLD engine please post an issue on GitHub, the feedback is really appreciated!

2. Several bugs were fixed including Issue 58.

MomentsLD was integrated by stanislavzlp and noscode .

1. We present a new engine - momi! It is now possible to perform demographic inference with momi engine as well as draw pictures of models with it. For more information please see documentation. If one will face problems with usage of momi engine please post an issue on GitHub, we really appreciate the feedback!
2. Option Recombination rate was added as future engines integrated into GADMA can reqiure it. ALso momi needs it to simulate data with msprime.
3. New input data - fastsimcoal2 input files. All three engines (dadi, moments and momi) are able to read this format now. For more information about format please see corresponding section in documentation.
4. GADMA is now available via conda (Bioconda).
5. We fix an error in a genetic algorithm that had ignored change of mutation rate and strength of GA. Unfortunately, the hyperparameter optimization that was performed before is no longer valid. We are going to rerun and obtain new hyperparameter values as soon as possible. Right now we change values of mutation rate and strength constants to default values.
6. The distributions of variable sampling were also updated: the time variable will be sampled from log normal distribution (was just normal) and option Random NA is False by default now. According to the experiments on several datasets GADMA with new distributions performs better.