Don't consider accepting on behalf of valid-but-accepted reviewers to be a validation error

Summary:
Fixes T12757. Here's a simple repro for this:

• Add a package you own as a reviewer to a revision you're reviewing.
• Open two windows, select "Accept", don't submit the form.
• Submit the form in window A.
• Submit the fomr in window B.

Previously, window B would show an error, because we considered accepting on behalf of the package invalid, as the package had already accepted.

Instead, let repeat-accepts through without complaint.

Some product stuff:

• We could roadblock users with a more narrow validation error message here instead, like "Package X has already been accepted.", but I think this would be more annoying than helpful.
• If your accept has no effect (i.e., everything you're accepting for has already accepted) we currently just let it through. I think this is fine -- and a bit tricky to tailor -- but the ideal/consistent beavior is to do a "no effect" warning like "All the reviewers you're accepting for have already accepted.". This is sufficiently finnicky/rare (and probably not terribly useful/desiable in this specific case)that I'm just punting.

Test Plan: Did the flow above, got an "Accept" instead of a validation error.

Reviewers: chad, lvital

Reviewed By: chad, lvital

Subscribers: lvital

Maniphest Tasks: T12757

Differential Revision: https://secure.phabricator.com/D18019