A couple of months ago, I have published a library libcerf that provides the missing functions cerf and cerfc, based on numerical code by Steven G. Johnson. Our implementation is accurate to 13-14 digits, which is good enough for almost every practical use - but in achieving this, one understands how much more work needs to be done to write an acceptable standard: it is not likely that this will be undertaken by anybody any soon.
So concerning your question about clgamma and ctgamma: don't wait for the standard. Search for code that just works. Ideally, wrap this code and provide a library like libcerf that is almost as good as a standard implementation.