notably, test normal + single + generated + Lua 5.3.5 only once,
then only run the runtime_test and compile_test for normal for all other permutations to help increase text matrix throughput
fix how `stack_aligned_protected_function` and its friends behave
add new internal handler details to allow for stack-based handlers with maximum performance
update `string_shim` typedef to simply be called `string_view` and use `string_view` in all public-facing APIs.
