Remove precompiled=no on windows.

[dhoegh]

As LLVM has been upgraded, is there any reason for not removing precompiled=no on windows? It reduces startup time and the test seem to pass on 64bit.

[vtjnash]

yes, we'll eventually do this. but there's a couple outstanding (upstream) bugs to deal with first (e.g. llvm3.5 didn't fix them as hoped). here's the quick list, if you're interesting in helping tackle them i'm happy to answer questions:

• the chief bug is patch llvm win64 dwarf4 relocations #15859
• we also need a custom SectionMemoryManager (that allocates all sections consecutively and near each other) and to patch llvm/lib/ExecutionEngine/RuntimeDyld/Targets/RuntimeDyldCOFFX86_64.h to use that image base from the memory manager (instead of hardcoding 0 to handle COFF::IMAGE_REL_AMD64_ADDR32NB) and a few other steps to hookup the .xdata section (basically, fix https://llvm.org/bugs/show_bug.cgi?id=24233). I think we might need to disable FORCE_ELF too (hopefully that doesn't break anything else, but I think we should be OK on that front).
• then we also need to set our llvm::Function unwind personality to call _seh_exception_handler (defined in libjulia).

[tkelman]

if it requires SEH, I'm guessing it'll require picking one of slower startup or worse backtraces on 32 bit?

[vtjnash]

win32 needs the first one, but not the others