Add support for the A extension

[Paul Donahue (pdonahue-ventana)]

Followup to https://lists.riscv.org/g/tech-debug/topic/76138908

I tried to provide flexibility for implementations that:

• don't return data on AMOs to x0 (and, for similar reasons, for loads to x0)
• have different ways of handling failing SC instructions
• do store data comparison in various places with relation to where AMOs are performed

[Ernie Edgar (ernie-sifive)]

👍

[Tim Newsome (timsifive)]

Content looks good to me. I just have some formatting feedback.

[Tim Newsome (timsifive)]

Thanks. github says there's a conflict in the PDF. Can you resolve that, so I can merge?

I think the simplest way is to merge master into this branch, resolving the PDF conflict in favor of master. But if you want to rewrite history so you never modified the PDF file in the first place that's also fine.

[Paul Donahue (pdonahue-ventana)]

Sorry. I always build the file to make sure there are no errors, to eyeball it, etc. I accidentally put it in one of the commits. I'll fix it.