Allow crucible_alloc_aligned to specify less alignment than default for type

[brianhuffman]

Currently crucible_alloc_aligned can only be used to specify a stronger alignment constraint than the default alignment for the type of the memory region. For example, a pointer to i32 has a default alignment of 4 bytes; crucible_alloc_aligned can be used to specify a stronger 8-byte alignment, but trying to specify an i32 pointer with 1- or 2-byte alignment will yield an error message.

We should relax this check, and allow things like crucible_alloc_aligned 1 i32 to specify a pointer to 32-bit integer with only 1-byte alignment.

LLVM allows us to write functions using unaligned read and write instructions. For example, we can have a function that takes an argument pointer of type i32* and then uses it with load statements that only assert 1-byte alignment (which might compile to special unaligned-read instructions on some architectures). Relaxing the minimum alignment check allows us to write accurate SAW specifications of such functions.