Proposal: Change @alignOf of bit-packed types to either return null or be an error

[rohlem]

Sibling-proposal, depends-on-but-half-in-conflict-with #7221 (read that one for context) .

Currently @alignOf returns a special value 0 for bit-packed types. The argument has been made that the definition of alignment requirements as a power of two is more consistent both semantically and for hardware, however this prohibits the special value 0.
Code dealing with byte alignment is (I think?) already not intended to be used with bit-packed types, so this may lead to mathematical errors in unsuspecting formulae. Instead, I think it's worthwhile to force a manual distinction, by either

• making @alignOf return an optional type.
  To not clutter code with .? (which isn't very intrusive in the first place imho) users can just use @alignOf indirectly via a helper function.
• or making @alignOf of bit-packed types a compile error (which would complicate comptime/reflection code using it though - you'd be forced to reimplement and perform the validity check in userspace).