probably need to change the behavior of revlanguage when the type of DAG node assigned to a variable changes

[mtholder]

Currently on development (and master), if you create a workspace variable with the = operator, and then reassign it as a constant DAG node, the behavior of the variable is as if it was still a workspace variable.

For example:

> a = 1
> str(a)

   _variable     = x
   _RevType      = Natural
   _value        = 1
   _dagType      = Constant DAG node
   _children     = [  ]

> a <- 1 
> b := exp(a)
> a <- 2
> a
   2
> b
   2.718282
> c <-1
> d := exp(c)
> c <- 2
> d
   7.389056

Note:

• a is represented as a "Constant DAG node", with no indication that it is a workspace var.
• applying the <- operator changes the value, but when it binds to a deterministic node (b) it acts like it was workspace variable.
• this means that reading a variable has side effects such that the results of scripts can vary depending on the order that they are executed.

Solution:

1. probably should report something in the str to indicate the status is a workspace variable
2. Either:
   * throw an error when reassigning type to a var (e.g. error on the a <- 1)
   * actually change the status of the node when reassigned.