closes #38

Replace the "use_for_plotting" column in mod$curves with the plot axis concentrations, which is far more useful.
Add "refence" argument to ec() and some minor improvements.

* Add plotting control for the observed points and improve the generation of the legend. See the `which` argument in `?plot_ecxsys`.
* Renamed `sys_tox_not_fitted` and `sys_tox_env_not_fitted` to `sys_tox_observed` and `sys_tox_env_observed` in the output of `ecxsys()`.

* The concentration vectors are checked for NAs, which raise an error when found.
* The use of `drc::ED()` was replaced with a custom inverse weibull function, which does the same but better.
* `predict_mixture()` accepts multiple values for the concentration of the second toxicant. Both concentration vectors must be the same length.
* `predict_mixture()` returns a data frame with the concentrations and effects. Previously it was only a vector of effects.
* `predict_mixture()` received a new argument "effect_max" which scales the returned effect values.
* Renamed the arguments of `predict_mixture()` to use underscore letters a and b instad of 1 and 2. For example model_1 is now model_a.

Add symmetry example to predict_mixture documentation and add license notice to every code file and to the readme.

* `ecxsys()` gains a new argument `curves_concentration_max` which sets the maximum concentration for the curves.
* `ec()` raises an error if the requested curve is too short to calculate the result.
* Improve merge request template.
* Update readme example.

Expand the capabilities of the plot functions, add `predict_mixture()`, and fix documentation of ecxsys() and predict_ecxsys().

Major rework of the source code and documentation. Refactor everything to improve usability and to make debugging and extending the package easier. See NEWS.md for details.