4D pack/unpack kernels

[cwpearson]

Interest expressed from Exascale Computing Project on 4D pack/unpack kernels. Need to create something substantially similar to https://github.com/cwpearson/tempi/blob/master/src/internal/packer_3d.cu. Furthermore, changes to make this function aware of the new packer:

tempi/src/internal/types.cpp

Lines 609 to 636 in fe16ae7

	std::unique_ptr<Packer> plan_pack(const StridedBlock &sb) {
	if (sb != StridedBlock()) {
	if (1 == sb.ndims()) {
	LOG_SPEW("select Packer1D for " << sb.str());
	std::unique_ptr<Packer> packer =
	std::make_unique<Packer1D>(sb.start_, sb.counts[0]);
	return packer;
	} else if (2 == sb.ndims()) {
	LOG_SPEW("select Packer2D for " << sb.str());
	std::unique_ptr<Packer> packer = std::make_unique<Packer2D>(
	sb.start_, sb.counts[0], sb.counts[1], sb.strides[1], sb.extent_);
	return packer;
	} else if (3 == sb.ndims()) {
	LOG_SPEW("select Packer3D for " << sb.str());
	std::unique_ptr<Packer> packer = std::make_unique<Packer3D>(
	sb.start_, sb.counts[0], sb.counts[1], sb.strides[1], sb.counts[2],
	sb.strides[2], sb.extent_);
	return packer;
	} else {
	LOG_SPEW("no packer for " << sb.str());
	return nullptr;
	}
	} else {
	LOG_WARN("couldn't plan_pack strategy for unknown type");
	return nullptr;
	}
	}