Pyungo

* Drop support / compatibility for Python 2.7.
* Inputs / outputs names are now optional:
python
graph.register()
def f_my_function(a, b):
c = a + b
return c

* Additional argument to Graph to prevent inputs deepcopy (Tosa95). This is useful when having large inputs when the deepcopy might be slow, but remove inputs immutability.

Allow optional input data for kwargs (veronicaguo)

1. Can pass list of `Input` / `Output` to `Graph`.
2. Names mapping.
3. Inputs validation with JSON Schema.
4. Some misc. refactoring and styling.

1. Refactoring, and introduction of Input and Output objects.

2. Contracts feature.:
python
graph.add_node(
my_function,
inputs=[Input(name='a', contract='>0'), Input(name='b', contract='float')],
outputs=[Output(name='g', contract='float')]
)

3. Nodes IDs refactoring (now uses uuid).

4. Run topological sort only when needed.

5. Everything accessible from pyungo __init__ (nelsontodd)

6. Docstrings / Sphinx doc / GitHub page

1. Fix "false parallelism" implementation. This require an optional dependency (`multiprocess` module) which is more friendly to work with stateful objects and multiprocessing.
2. Ability to pass predefined inputs at the node definition:
python
graph.register(inputs=['a', {'b': 2}], outputs=['c'])
def f_my_function(a, b):
return a + b

1. Nodes can be run in parallel using Python multiprocessing module:
python
graph = Graph(parallel=True, , pool_size=3)


2. `inputs` and `outputs` parameters are mandatory and will raise an explicit error when missing.

3. Add a longer test using `pvlib`