UDF support in QL #558

See #567 for a logical prerequisite that makes this feature possible.

The last remaining disparity between Exp API and QL is the lack of user-defined functions in the latter. Let's create a syntax to pass UDFs as QL parameters that would to emulate mapVal(..) and mapBoolVal(..). Some examples...

1. Anonymous functions passed as positional parameters. Easy to define on the spot, but lack types. ?f is a special parameter syntax that expects a function:

df.cols().select(
   "?f(a, b, c, d) > 5 and bool(?f(e, ?))",
   UdfN.of(this::op1),
   Udf2.of(this::op2),
   100);

2. A context for named and typed functions. More effort to setup, but the functions are reusable and the expression syntax is clean:

// Suppose we can even stick one of these contexts in the shared 
// Environment to define global functions
ExpContext context = ExpContext.of()
        .intFunction("myOp", this::op1)
        .boolFunction("myPredicate", this::op2);

df.cols().select(
        context, 
        "myOp(a, b, c, d) > 5 and myPredicate(e, ?)", 
        100);