Improve string parsing on targets that use 32-bit pointers but also have fast 64-bit integer arithmetic, such as aarch64-unknown-linux-gnu_ilp32 and x86_64-unknown-linux-gnux32 (#1182, thanks @CryZe)
Optimize string parsing by applying SIMD-within-a-register: 30.3% improvement on twitter.json from 613 MB/s to 799 MB/s (#1161, thanks @purplesyringa)
