Support allocators for STL containers owned by xarray and xtensor

[llllllllll]

Adds the ability to specify an allocator type for the elements and shape of an xarray and the ability to specify and allocator for the elements of an xtensor

[SylvainCorlay]

Thanks @llllllllll this is definitely something we want to allow, also for the SIMD acceleration that we will be enabling soon, which requires the aligned memory allocator implemented in https://github.com/QuantStack/xsimd for better performance.

The failing windows test seems to be legitimate.

[llllllllll]

I will look into the windows case sometime this week. Is there a good way to get access to the microsoft compiler on a gnu+linux system to test against?

[JohanMabille]

@llllllllll I don't know about that, I have both systems installed. I'll have a look on this.

[JohanMabille]

@llllllllll Adding the namespace xt before fma to disambiguate the call fixes the problem. I think we can use that workaround for now, and dig deeper into this issue later; I'll let you fix it, opening a PR on your PR for such a change is a bit of overkill :)

[llllllllll]

Thank you for looking into this!

[JohanMabille]

No problem, thank you for implementing this feature!