Revert "windowed post generation now returns faulty sectors"

[magik6k]

Reverts #12

It doesn't really make sense to treat skipped/faulty differently

[whyrusleeping]

doesnt it? don't we want to use this information to adjust our next call to generate windowed post?

[anorth]

I agree with @magik6k, I think you need to know which sectors are going to fail before calling this, so you can then rework the sector infos to mask out the faulty sectors and try again.

I would propose a flow where:

• this method fails quickly (after sampling) if any sector can't be challenged, and returns the failing sector ids or indices
• the mining worker then adds them to the fault set, recomputes the arguments to this function, and tries again
• after success, the mining worker propagates the skipped sector ids

[Stebalien]

At the moment, skipped/faulty are the same thing. We can revisit this if that changes, but this patch is going to lead to interface churn for no good reason at the moment.

[Stebalien]

(merging because this is blocking other work)