ACLE/SIMD32: add ssub8 and usub8

[paoloteti]

• add ssub8 and usub8
• bump instruction limit to 29

cc @japaric

[paoloteti]

Travis is green. AppVeyor failure is unrelated and looks like a well known issue.

[gnzlbg]

@paoloteti heads up: I'm waiting for this PR upstream to be merged before adding more features to stdsimd: rust-lang/rust#58675 since we need that to port libstd to the 2018 edition. Once that is done, I'll start merging other PRs as well (feel free to submit more PRs implementing the other APIs or to add those to this one in the meantime).

Also I would prefer if someone more familiar with ACLE like @japaric@Amanieu or @parched would review the PRs related to these.

[Amanieu]

LGTM

[gnzlbg]

Thanks!