[PATCH] Implement aarch64 NEON copy_lane intrinsic with float type

[Tim Northover]

Hi Kevin,

The Clang patch looks good, and as far as I can see the LLVM one is OK
except that it needs CodeGen tests.

Oh, and the spacing on this line and the int version is wonky:
+        EltVT = (EltSize == 64)? MVT::f64: MVT::f32;

Cheers.

Tim.