Core: Add restricted_dumps helper

[duckboycool]

What is this fixing or adding?

Adds a restricted_dumps method to Utils corresponding to restricted_loads which errors if the object cannot be unpickled with restricted_loads, and replaces some current uses of pickle.dumps with it. I'm not positive if some of these uses might be safe to not have the check, although they might also be able to remove the corresponding use of restricted_loads if so. There seemed to be no relevant uses of pickle.dump to a file instead, but it might be better to implement a RestrictedPickler class instead of just the function if the flexibility is desired.
I also changed the test_pickle_dumps test to also try getting the default with from_text since this value might be different than the one gotten with from_any(option.default) and cause issues.

How was this tested?

I tested the changes with a problem that was present in the satisfactory world options that caused it to error when generating a single player game from the options page. It fails immediately in this case instead of first trying to generate and then failing later when unpickling. I also ran the updated test on the world and it was able to find the error whereas the world passed the test before.

If this makes graphical changes, please attach screenshots.

🚫🥒

[NewSoupVi]

Makes sense to me

[NewSoupVi]

I'm gonna merge this at the start of next release cycle

[beauxq]

looked over code
local generated, served, and played a game

[Berserker66]

This does make dumping quite a bit slower, and in some situations I feel it should never fail, like webhost upload. - where the increased performance cost would also be among the highest impacted.

[duckboycool]

I think the two in process_multidata (and the TestTracker setup) should definitely be safe since they came from a call to decompress which uses restricted_loads in the first place. The ones on multisaves also seem like they could be fine. Do you have an idea about whether some restricted_loads calls might also be unneeded?

[beauxq]

This does make dumping quite a bit slower, and in some situations I feel it should never fail, like webhost upload. - where the increased performance cost would also be among the highest impacted.

I don't know which serialization you mean by "webhost upload".
Can you point out which restricted_dumps calls in the diff you're most concerned about for performance?

We've seen it fail in at least 3 of these places. But, yeah maybe some of them don't need it.