Enable load cast `xt::xtensor<...>`

[tdegeus]

Sometimes my API simply does not permit easily to template, or I just want to wrap quick and dirty, or on needs to work around a bug. For these case I still want to be able to wrap

void foo(const xt::xtensor<double, 1>& a);

Currently this is explicitly prohibited in xtensor-python

xtensor-python/include/xtensor-python/xtensor_type_caster_base.hpp

Lines 77 to 80 in 9aa58f8

	bool load(handle /*src*/, bool)
	{
	return false;
	}

I find this too restrictive, and I think that xtensor-python should have a way to do it, for instance by setting a flag or include a header. We should give the possibility to work around something, even if it comes at the cost of a copy.