Run initial SIMD spec tests

[abrown]

This change will allow us to verify that the SIMD implementation matches what is expected in the spec tests published to https://github.com/WebAssembly/simd/tree/master/test/core/simd. Some notes:

• it depends on the changes in Support running spec tests in wasmtime cranelift#992; when these are published as a new cranelift version, 7848f68 should go away
• it depends on an updated version of wabt-rs to fix a SIMD parsing bug; once Update wabt dependency to latest HEAD pepyakin/wabt-rs#56 is merged, 07541d0 should go away
• it adds a way to run a single spec test in build.rs and enables SIMD for spec tests
• it adds SIMD support throughout wasmtime
  This is a work in progress.

[abrown]

@yurydelendik noted that wasmtime-debug might also need support for v128: e.g., https://github.com/yurydelendik/wasmtime/blob/853238529e4df416e36a364c7df99019c3cabdd1/wasmtime-debug/src/transform/simulate.rs#L162

[sunfishcode]

I forget where you asked me this, but to your question about exposing Uimm128:

Stepping back and looking at it, Uimm128 is actually misnamed, because it's not a scalar integer immediate, like the other Uimm* types are. (We should perhaps rename them too, as Uimm* isn't super clear, but that's a separate issue). If we renamed Uimm128 to something like V128Imm, I think it'd be fine to expose that in APIs. And it could have a Display implementation so that you don't have to use {:?} as the PR currently does.

[abrown]

I guess global get/set is still not working...

[abrown]

This is almost there; only bytecodealliance/cranelift#993 needed to address the failing address.wast spec test:

...
@0267 [RexOp1rout#01]               v7, v8 = iadd_ifcout v1, v6
@0267 [-]                           trapnz v8, heap_oob
...
spec_testsuite/address.wast:3: WebAssembly compilation error: Compilation error: Verifier errors:
- inst8: has an invalid controlling type iflags
- inst8: Trapping instruction must have an encoding

[sunfishcode]

Looks good!

[abrown]

@sunfishcode thanks for all the debugging help!