Project:LLVM/Split packages

Starting with LLVM 4.0.0, various LLVM projects are available as separate packages again.

Why split?
LLVM is commonly built as a single big CMake tree with the top-level (LLVM) CMake scripting finds tools in subdirectories and builds them. However, in 4.0.0 the upstream build systems were fixed to the point where building tools separately became possible.

Split builds mean that instead of one big package with multiple USE flags, we have multiple packages that can be built separately. This has a number of advantages:
 * It is possible to install/remove additional LLVM projects without rebuilding the remaining ones. For example, if you want to try LLDB, you don't have to rebuild LLVM and clang.
 * The build trees are smaller, and builds of single packages are shorter. As a result, building LLVM takes less space and retrying on failures is easier.
 * The test suites are run separately, so a test failure in one package does not cover the result from another.
 * It is cleaner as to which components USE flags apply, and e.g. which projects are multilib.
 * The ebuilds become simpler, e.g. because there is no need to apply multilib logic with a lot of conditionals for non-multilib parts.