Servo's compile time went down by about 20% or so. We think that it's because non-zeroing drop enabled LLVM's memcpy optimizer (which is currently pretty fragil

Servo's compile time went down by about 20% or so. We think that it's because non-zeroing drop enabled LLVM's memcpy optimizer (which is currently pretty fragile pending the MemorySSA rewrite) to do a better job.