[RFC] Replacing our regex implementation

[Gabor Kovesdan]

Thanks to all for the valuable comments.
I've decided to continue the work with TRE because
- Many people don't like the idea of adding C++ code into libc
- I'm not very skilled in C++

However, in a later phase of development, I think re2 may be a good 
source of ideas in optimizing the performance. For now, I think if the 
new code without optimizations is at least as fast as the old regex 
code, that's fine. Imho, the motivation of replacing it is not just the 
performance but also the fact it is outdated code and not maintained any 
more and also the lack of wide character support. Once it is 
feature-complete to replace the old code, we can start thinking of 
optimizations. And as proposed, I'll separate those from libc for now 
and later I can propose some changes to the TRE maintainer.

Gabor