[PATCH] Reload undefined atoms to use possibly new undefines.

[Shankar Kalpathi Easwaran]

  Do you think that we should revert fallback until lld models COFF external weak symbols(mainly to avoid confusion in the resolver). I still think that Alias is a better option here(the model which Nick suggested in early emails). What do you think ?

http://llvm-reviews.chandlerc.com/D2162