Refactor internal error handling

[Rua]

This renames the internal Error type to VulkanError (so that it no longer conflicts with std::error::Error) and auto-generates it from vk.xml. The check_errors function and the Success enum are also removed. There is also a new impl From<ash::vk::Result> for VulkanError, which is tolerant to error codes it doesn't know about; it will wrap them in the Unnamed variant, whereas check_errors would panic before.

My intention for the future is to return VulkanError directly to the user. Currently, there are From<VulkanError> implementations for every error type, which translate known errors for that function. But this adds a lot of duplication, especially for each Display implementation that has to repeat the same error messages for standard Vulkan errors.

Moreover, these From<VulkanError> implementations currently panic if a Vulkan API function returns an error that is not expected for that function. Some Vulkan implementations will return errors that aren't in the spec, and Vulkano will therefore indiscriminately crash the user's program based on something they have no control over, which is not very helpful. So instead of trying to translate each known error and panicking on the remainder, returning VulkanError directly would be more tolerant of unusual Vulkan implementations, letting the user deal with the unexpected behaviour instead.

[AustinJ235]

Awesome!