Optimize structure packing within the SDK

[andy31415]

#36995 did some experimentation and there are definately savings to be had, however those can also be at the cost of readability/logical grouping. We have to see if any savings exist that do not impact maintainability (and those can just be applied) or if some changes are worth it enough to make anyway.

To check structure packing, pahole can be used (note though that across different architectures we get different results ... we should probably focus on arm32 as the typical small target):

e.g. pahole --holes=1 --show-private-classes chip-qpg6105-lighting-example.out or pahole --show_private_classes ./out/qpg-qpg6105-light/obj/third_party/connectedhomeip/src/app/libCHIPDataModel.ReadHandler.cpp.o