Consolidating error management

Makes "refCountedPayload" nothrow when
``autoInit == RefCountedAutoInitialize.no``

Also makes Payload.isInitialized nothrow

Big Changes for Array:

First: I added some new operators:
The range wide opUnary and opOpBinary operators, that allow doing an operation on all elements. This was added on Array.Range only (not on Array itself): This is to stay close to native array's behavior:
a += 5;//error with both native arrays and Array
a[] += 5; //OK with both native arrays and Array

Second: Added an invariant: Extracting a range from an Array forces that array's initialization. This is actually VERY valuable, and aids us in a twofold manner:
#1: Performance: None of the range's operations need to validate that _data is initialized.
#2: Provided the *assert* "_b <= _outer.length", we are guaranteed that "_outer._data._payload[_a .. _b]" is valid, making for quick and easy slicing.

Second: A *big* change in philosophy regarding bounds checking: First, both Array and Array.Range implement a private "get" method that always returns a valid raw array (see implementation for details). This has two advantages
#1: Centralized logic: Once we have a valid slice, every "op" implementation become trivially simple, and become nothing more that a forward to get's range:
#2: Since we are operating on valid slices, we can rely on the language's array bounds checking for bounds enforcement: This actually makes the error messages more verbose, while costing less in developer code, executable size and run-time.