Refactor AllocationCreateInfo

[marc0246]

This refactors AllocationCreateInfo such that it can be used together with BufferCreateInfo and, in the future, ImageCreateInfo, to construct buffers and images, removing BufferAllocateInfo. Doing this before users start using this API in 0.33.

What is still unresolved is whether ignoring BufferCreateInfo::size when in these constructor functions is the right way.

Changelog:
Add:

### Breaking changes
- `AllocationCreateInfo::{requirements, allocation_type, dedicated_allocation}` fields were removed.
- Added the parameters `requirements`, `allocation_type`, `dedicated_allocation` to `MemoryAllocator::allocate[_unchecked]`.

[Rua]

I see two other options for size:

• Require it to be always zero.
• Pass it on to the buffer creation when it's not zero, allowing the user to override the size if they want.

[marc0246]

If the user can override the size, then that would cause issues if it's smaller than required for T, or [T] given some len. And in the case where the size is larger, that would result in the user not being able to access the rest of the bytes in those cases. Therefore I think that the only viable option is nr 1, at least for now.

[Rua]

Let's go with that then. I think requiring a fixed value, especially one that is otherwise disallowed, is better than just ignoring the value. It allows us to change our mind later.

[marc0246]

Yes, I very much agree. That's what is currenly in the code, except that the requirement for it to be zero is only a debug_assert. Do you want me to make it stronger, and/or anything else?

[Rua]

Yeah I think a full assert would be better.

[marc0246]

Alrighty done.