It's not possible to share them across threads, so in case we decide to use multithreading in the future, it's best to mark them as thread_local right away, even if it's a no-op right now.
Looks like we've been stuck on Emscripten fastcomp backend, misssing out on all new optimisations between LLVM 6 and LLVM 11 and going via slow asm.js pipeline when building.
This changes Docker images for Emscripten projects to point to emscripten-upstream instead and commits the updated artifacts.