Lint against redeclared variables

[gaearon]

There was recently a regression that caused variables to get redeclared: https://github.com/facebook/prepack/pull/2255#issuecomment-405923183.

To catch issues like this, I'm introducing no-redeclare as an additional lint rule that runs on serializer tests. I'm also adding a way to disable individual lint warnings since there's a case where we intentionally test code that's not-so-great.

This is still not sufficient because we do emit redeclaring variables. So this PR is expected to fail the tests until these cases get fixed.

I don't know if it's a recent issue or not, but I think it needs fixing. So I'm sending this PR as the first step to surface these issues. Once they get fixed, we can merge this.

[NTillmann]

It's an issue. It's likely because generators currently do not form a tree, but a DAG. Until that's fixed, redeclared variables are to be expected.

[gaearon]

Would it be useful to add lint suppression to known cases where it happens and hope to reduce them in the future? Or is it too non deterministic?

[gaearon]

It looks like this fails on a lot more tests then the regressed NestedConditions.js I added. I’m guessing for different reasons. Is this expected?

Not by me, but yes, seems like it's an unexpected existing flaw

[trueadm]

Fixed conflicts from master. Let's try and get this one in if possible :)

[gaearon]

@trueadm You'll get it in? I forgot why we didn't last time.

[trueadm]

@gaearon Yep, the tests pass now.