Disable Recommender Experiment, make original job graph the default

[mrdavidgagnon]

The experiment preferred the static job graph.
The experiment found the original job graph performed better. Let's stick with that while we queue up a new recommender algorithm

[LswaN58]

Updating this to-do, since it doesn't look like original change was made yet.

We discovered that the numbers were backwards in original analysis of the experiment, and the original job graph actually had best performance in terms of players completing more jobs. So we should stick to the original job graph.

Still turning off the recommender system, of course.
In general, just turning off the experimental conditions to go back to the original.

[LswaN58]

@LswaN58 test this. I believe this is also under play/wake/ci/develop.

[LswaN58]

@BeauchesneFieldDay @jmmathew-cpu

I've tested this. Looks like the logging is reporting the branch as develop, not one of the experimental versions. Also:

1. I didn't encounter any "you're not ready for this job" recommendations, while following the exact same sequence that gave me such a recommendation when I tested on staging (where the fail-prediction system is still enabled). Seems safe to say this is fully turned off.
2. I don't have any way to directly validate that the old job graph is in use, but I didn't notice any differences in what was unlocked compared to staging (where the old job graph is explicitly listed in the branch name).

Based on these points, I think we're good to go!