Align the start of stackslots instead of the end

[tertsdiepraam]

Hi! I think that the stackslots weren't properly aligned, because the padding for the alignment is being added to the end of the slot instead of at the start, meaning that the slot itself isn't really being aligned.

This was creating issues for me where I requested a slot that was aligned at 16 bytes, but in reality it wasn't.

With this change this case from the tests:

ss0 = explicit_slot 8, align=16
ss1 = explicit_slot 8, align=16
ss2 = explicit_slot 4
ss3 = explicit_slot 4

gets these offsets:

ss0: 0
ss1: 16 (because size of ss0 + padding to get to 16)
ss2: 24 (because size of ss1)
ss3: 32 (because size of ss2 + padding to get to the word size of 8)

Note that the align=16 of ss0 is essentially a noop now.

And this case from a test I added:

ss0 = explicit_slot 8
ss1 = explicit_slot 32, align=16

gets these offsets:

ss0: 0
ss1: 16 (because size of ss0 + padding to get to 16)

I hope I didn't make any mistakes here or misinterpreted what stackslot alignment should do. Let me know if I need to change anything!

The issue was originally found here: https://bytecodealliance.zulipchat.com/#narrow/stream/217117-cranelift/topic/Retrieve.20alignment.20of.20architecture/near/471006324

[cfallin]

Thanks a bunch for fixing this!

[tertsdiepraam]

Thanks!