Align allocWorkBufffers to 32 bytes

[jcberentsen]

Possible fix for #53

[jcberentsen]

This particularly fixes the original #53 for me. Perhaps there are other places that should have the same alignment?

[newhoggy]

I'm inclined to think that if this works for you then that's all that is needed. The alignment is only needed when passed to C functions that make use of the SIMD instructions that require the alignment.

[newhoggy]

Can you check if the fix in this PR works?

#58

I prefer this version since the 32 isn't repeated reducing the opportunity for error.

[jcberentsen]

Testing #58 works good on my end. I'll close this PR
Thank you for taking the time on this issue!

My usecase went from 3s down to 0.58s by computing the SIMD cursor!
The hw-json project is amazing!