Silent casting of int symbol's state in some cases

[JoelPasvolsky]

Silent casting of int only when boundaries are defined:

1. Got what I expected:

   >>> i = model.integer()
   >>> with model.lock():
           model.states.resize(1)
           i.set_state(0, -2.5)
           print(i.state(0))
   ValueError: Invalid data provided for node

2. Got surprised:

   j = model.integer(lower_bound=-5, upper_bound=0)
   >>> with model.lock():
           model.states.resize(1)
           j.set_state(0, -2.5)
           print(j.state(0))
   -2.0

[arcondello]

Also needs a better error message.

[arcondello]

That said, I think we want to silent cast in both cases. To follow NumPy

In [11]: arr = np.zeros(1, dtype=int)

In [12]: arr[0] = 2.5

In [13]: arr[0]
Out[13]: 2