simd128 opcode update and new instructions

Need to catch up with the state of `simd128` WebAssembly proposal, though it might make sense to wait until after the opcode renumbering - WebAssembly/simd#209.

There is a number of new operations, in particular `swizzle` had been reintroduced, but with variable mask; general case would be expensive to implement without `phufb` support, but we have part of 4-lane variety still in the sources (albeit with a constant mask).