TabManager.TabRanker.Result is always "Success" and should be replaced with a DCHECK.

The TabScorePredictor has to load the model preprocessor config (once) and then pre-process each input before running it through the inference function. Theoretically this could fail if the config failed to load or if the input wasn't formatted properly, so the TabScorePredictor currently emits an UMA histogram with the success or error result of these operations.

Out of >100 million entries so far, all of them have been "Success". This confirms that these operations shouldn't fail; a failure would be the result of a coding error. We should remove the TabManager.TabRanker.Result histogram and instead enforce success with a DCHECK.