partial optimisation for ord(Py_UNICODE) - better optimisation that returns a C integer result requires type analysis refactoring

support redundant parsing of string literals as unicode *and* bytes string, fix 'str' literal assignments to char* targets when using Future.unicode_literals

Also expand comments slightly.

implement C array iteration with explicit step size (#574), always coerce start/stop offsets in C array iteration to Py_ssize_t

compile is necessary.

We really need to make these enabled by command line flags...