Add DiskIo and DiskIo2 protocols

[e820]

Adds the EFI_DISK_IO_PROTOCOL and EFI_DISK_IO2_PROTOCOL protocols from the UEFI specification that build on top of the block I/O protocol to provide byte-oriented disk I/O.

[nicholasbishop]

Thanks for the PR!

I think the DiskIo part looks good, but we should consider the safety of the DiskIo2 interface. Since the operations are asynchronous if an Event is passed in via DiskIo2Token, the buffer and token parameters needs to remain valid for the whole lifetime of the request, but this isn't currently enforced. I'm not quite sure what a good safe interface would look like yet. As an intermediate step, we could change the current functions to be unsafe, maybe also rename them to something like read_disk_raw/write_disk_raw, and change the buffer and token parameters to be raw pointers. Alternatively we could try to jump right into figuring out what a safe interface looks like, I'm interested to hear ideas.

Also, it would be good to add some tests for the new protocols. I think these could go in uefi-test-runner/src/proto/media/known_disk.rs.

[e820]

DiskIo2's read_disk & write_disk were made unsafe and renamed. I also added a small test for both DiskIo & DiskIo2.

[nicholasbishop]

Thanks for the updates and added tests! I added some comments on the tests.

[nicholasbishop]

In trying to track down why the Event never gets signaled, I realized that this struct does not currently have the correct layout. We aren't using NonNull in Event like we do with Handle, so Option<Event> doesn't get the same layout as a raw pointer.

I filed #477 to fix this, but for now this line should be changed to pub event: Event.

[e820]

I think in this case this actually isn't necessary because I interpreted the specification wrong. The spec talks about the token "field" to be nullable, but it seems what it means is that the token argument to ReadDiskEx/WriteDiskEx can be NULL - meaning I could simply change the parameter type.

[nicholasbishop]

Looks good, thanks!