bug(nimbus): Prevent mismatches between NimbusReviewSerializer and the front-end allowing invalid experiments

[brennie]

Because:

• #112832 added the segments field to the NimbusExperiment model.
• It did not add it to the NimbusReviewSerializer, which caused the serializer to infer a default serializer for the field which marks it as a required field.
• Therefore, the NimbusReviewSerializer was always throwing before we hit its validate() method (due to the invalid segments field), which prevented the majority of the validation from running.
• And the front-end was not checking the ready property from the useReviewCheck() hook, which allowed launching invalid experiments.
• Our tests were always creating NimbusExperiments with non-empty segments fields.

This commit:

• Adds a segments field to the NimbusReviewSerializer which makes the field optional.
• Adds a test that explicitly creates a NimbusExperiment with an empty segments field and verifies it does not raise a ValidationError. (This test fails without the changes to the serializer.)
• Changes our NimbusExperimentFactory to randomly use between 0 and 2 values for segments, primary_outcomes, and secondary_outcomes so that we have broader test coverage over all possible inputs.
• Updates our front-end to use the ready key instead of invalidPages to prevent launches.
• Updates the front-end to instruct users to come to #ask-experimenter if ready is false but there are no invalid pages (i.e., they've hit a programming error).

Fixes #11767

[jaredlockhart]

Okay I tested everything locally, all looks and works good!

Thanks for catching and picking this up so quickly @brennie 🙏 🙏 🙏