Fix the test runner

[gsnedders]

After the manifest changes in #125 and #90 landed, the runner broke. We should fix this before updating the tools submodule in wpt.

[jeremypoulter]

I started having a look at this, but would be helpful if you could summarize what changed and what is the scope of the required fixes?

FYI so far I have fixed an issue with update_manifest.py not running.

[gsnedders]

There were a bunch of changes to the manifest format, so all the changes should essentially be handling of the JSON file in runner.js (along with the update_manifest.py changes which you've already done).

The big things are the dropping of local_changes (so all the JS code related to that can simply be killed!) and the change to each item being an object of path -> tests from being a list of tests which should have no effect on the JS code (you just ignore the path). The from_json methods in manifest/manifest.py and manifest/item.py should be taken as the semantics of the JSON.

Really here the scope is to just get it working as well as it did before, and we're unlikely to require tests for this (because it's so much more of a pain to test and there's no infrastructure set up for tests for the runner yet!).

[jeremypoulter]

Thanks, I will have a look at this, FYI my working fork/branch is https://github.com/jeremypoulter/wpt-tools/tree/test_runner_fixes