Add multiply-add / multiply-sub (MADD / MSUB) operations. Also implemented sub / add wasm versions

[MaxGraey]

[dcodeIO]

Given that mul is already performed in Wasm and that add / sub aren't so slow, and given the scope of this package, I'm wondering if extending the API with rather uncommon methods as proposed here is worth it?

[MaxGraey]

mul + add pretty common operation which could be speedup by calling only once instead two (which also required unpack same arguments twice). For example. I could remove rest wasm methods like msub, add and sub which added with this PR if you don't see any sense for it.

[dcodeIO]

Having a feeling that peak perf isn't the right motivation in context of this library. In practice it still emulates so fusing two ops seems neglectable?