Layout test editing/spelling/spelling-huge-text.html tests performance rather than functionality.

We should convert it to a performance test.

We already have good UMAs for a finch trial:
- SpellCheck.api.async: number of characters in each spell check request
- WebCore.ScriptedIdleTaskController.IdleCallbackOverrun: how long an idle task runs overtime

So the only remaining task is to implement performance benchmarks. And the priority is lower because the benchmark does not block the launch process: we only need to know the performance gain from benchmarks.

Whether the performance gain comes at a price will be found out from finch trial.