The 2017 model appears to perform well with Blacklist Override enabled, so we'll be rolling it out in place of the faulty 2018 model.

To make sure that larger population don't exhibit poor performance, we'll progressively rollout the 2017 model with and without the blacklist override in parallel and compare the data before deploying the blacklist override to 100%

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/22981ad9808e8d7b801ea913bbe512af779af69e

commit 22981ad9808e8d7b801ea913bbe512af779af69e
Author: Anthony Vallee-Dubois <anthonyvd@chromium.org>
Date: Tue Aug 21 16:54:55 2018

[Language] Default Translate Ranker Model

1/ Add new Ranker config to field trial testing
2/ Set the default model to be consistent with the 99% Finch group

Bug: 852921
Change-Id: I04497f6a80d810a506e40b7ff9e4d1500ffc191c
Reviewed-on: https://chromium-review.googlesource.com/1179814
Reviewed-by: Robert Kaplow (slow) <rkaplow@chromium.org>
Reviewed-by: Michael Martis <martis@chromium.org>
Commit-Queue: anthonyvd <anthonyvd@chromium.org>
Cr-Commit-Position: refs/heads/master@{#584791}
[modify] https://crrev.com/22981ad9808e8d7b801ea913bbe512af779af69e/components/translate/core/browser/translate_ranker_impl.cc
[modify] https://crrev.com/22981ad9808e8d7b801ea913bbe512af779af69e/testing/variations/fieldtrial_testing_config.json