More static analysis during DAG instantiation

[thorwhalen]

We'll use the following functions:

def foo(a, b):
    return a + b

def bar(x, y):
    return  x * y

binding to a FuncNode name (id)

The following makes a valid DAG, but it's probably not the one the use wanted.
Here, the dag acts as if there was an extra root var node 'foo' and computes from there.

from meshed import DAG, FuncNode
from i2 import Sig

dag = DAG([
    FuncNode(foo, name='foo'),
    FuncNode(bar, bind={'y': 'foo'})
])
assert str(Sig(dag)) == '(foo, x, a, b)'  # ?!?!
assert dag(9,1,2,3) == (9, 5) 
dag.dot_digraph()

binding to a function

dag = DAG([
    FuncNode(foo, name='foo'),
    FuncNode(bar, bind={'y': foo})
])
# ValidationError: Should be a str: <function foo at 0x121db68b0>

Probably what the user would want in the two cases above is binding the y param to the output of a function.
We could do this automatically for the user.

[thorwhalen]

Still reproduces on master (0cb2609):

```python
from meshed import DAG, FuncNode
def foo(a, b): return a + b
def bar(x, y): return x * y

dag = DAG([FuncNode(foo, name='foo'), FuncNode(bar, bind={'y': 'foo'})])
str(Sig(dag)) # '(a, b, foo, x)' -- foo (the FuncNode's name, not its out) is an extra root var
dag(9, 1, 2, 3) # (10, 30) -- computes foo(a,b) separately AND takes a disconnected 'foo' root input
```

No validation currently catches "you bound to a FuncNode's name instead of its out" -- it silently creates a same-named disconnected root var instead. Leaving open; a real fix here is a validation-time design choice (error? warn? treat name-binding as an alias for out-binding?) rather than a one-line patch.