Paramtools

- Promotes the tax parameters to be the prominent example in the documentation (35).
- Behavior-Responses parameter example and cleaner error messages (37).
- Add ability to delete parameter value objects by setting their value to `None` (38).
- Fix a bug where a parameter name specified in the defaults collides with a name used by the `Parameters` class (39).
- Use conda to set up the paramtools development environment and for packaging (41).

This release improves the format of the README files and adds fields and validator documentation.

ParamTools 0.3.2 contains a bug fix that specifies that the data files should be included in the ParamTools package.

ParamTools 0.3.1 contains a bug fix that stores the example JSON config files in the ParamTools package. This way, the examples work correctly without extra configuration.

ParamTools 0.3.0 enables users to more easily access the values of each parameter as NumPy arrays. The shape of the array is determined by the validators set in `schema.json` and the state of the `Parameters` instance which can be set via `Parameters.set_state`. The final piece was to provide a flag `array_first` which allows the user to specify whether the values should be accessed as NumPy arrays by default or not.

- ParamTools now requires NumPy to be installed. (21)
- Dimension validation information in `schem.json` is used to define the default array shape for each parameter's value objects. (22)
- `paramtools.Parameters` now has a "state" concept which is used to activate parameter values that lie in the cross section of the dimensions specified with `Parameters.set_state`. (23)
- `paramtools.Parameters` has an array first option which allows the user to access the parameter values as arrays by default. (24)

Install with: `pip install -U paramtools`