Add typedef for anonymous structures

[X-Ryl669]

See #3
I'm trying to create a function signature from crefl database. Since it's not possible to refer to an anonymous structure in a function signature, we need the typedef as written by the initial code.

This adds the feature, by substituting an anonymous struct to a typedef if possible.
However, it breaks tests, but I don't know if it's expected or not.

[michaeljclark]

thanks, I'll take a look. I'll need some time to inspect the resulting graphs. I haven't yet automated the testing beyond creating text output that I diff locally. that of course does not handle structural changes that require changing the baselines, like referencing the most specific type in a decl versus a desugared type. as I mentioned before, the tests are not buttoned down in CI because the graph schema is still under some flux. where I got up to was function addresses, libffi integration, qualifiers and dependent expressions for VLAs using operator patterns.

C++ is in the back of my mind but I wanted to complete C first because while I'm building this for use in a C++ app, most of the reflection stuff that I need for an entity-component system is simply structures, functions and intrinsic types. I do need a C++-style invocation on an interface, but I need to build it in terms of C so that C can call C++ virtual functions using crefl and libffi.