Upgrade to the latest LLVM

[langston-barrett]

cclyzer++ has only been tested with LLVM 10. Upgrading would involve:

• Reading through the LLVM changelogs for any relevant changes in semantics
• Upgrading the FactGenerator and C++ interfaces to the new LLVM API

[langston-barrett]

We should at least look into supporting multiple versions of LLVM. This makes cclyzer++ applicable in a wider variety of situations (e.g., a project where an older version of LLVM is required). More prosaically, it means we can keep the cclyzer++ submodule in the MATE project up-to-date without spending time upgrading MATE's LLVM support. This is valuable because MATE has a number of tests that could help find bugs in cclyzer++.

That being said, there might be major downsides to supporting multiple LLVM versions, the most likely candidate being an inordinate expense in terms of developer, build, or CI time; or in terms of (Docker) build complexity.

[langston-barrett]

Reopening because LLVM 15 isn't fully supported yet due to opaque pointers.