Tup and C++ module support

[Ben Boeckel]

Hi,

There has been discussion with the C++ committee about how to build
modules since it generally requires more smarts from build systems.
Basically, `g++ -c foo.cpp -o foo.o` is no longer trivially
parallelizable and compilations must be done in order based on the
contents of the compiled files. I was wondering what kind of approach
Tup would be using for this since it is different in the build system
space than tools like make and ninja with its strace-style detection.

The latest thread on this is here:

http://www.open-std.org/pipermail/tooling/2019-March/000609.html

Discussion is here:

http://www.open-std.org/pipermail/tooling/2019-March/000520.html

Here is our paper to the ISO C++ committee on how CMake's approach to
building Fortran code is applicable to C++ modules:

https://mathstuf.fedorapeople.org/fortran-modules/fortran-modules.html

Thanks,

--Ben