perf(specs): Clean up factories usage in spec/specs in directories k-o

[vankiru]

Context

Currently, around 50% of the spec runtime is spent preparing data and creating factories. A lot of this data is either duplicated or recreates the same objects multiple times, so we can get rid of a good chunk of it and significantly speed up the specs.

To avoid creating unnecessary factories, we can use let_it_be, before_all, and create_default from test-prof. More info here.

Description

This PR cleans up the services specs in folders k-o and removes unnecessary factory usage.

Here are the numbers before and after the cleanup. The times below are for running the specs in a single process:

Metric	Before	After
Total # of factories	97,841	63,994
Time spent creating factories	11:51.790	07:40.512
Total time	18:00.870	14:22.880

Overall results

This PR is part of a series of PRs focused on cleaning up factory usage across different specs.

Once all the changes are merged, we expect to:

• Reduce the number of factories by ~50k
• Cut spec runtime by around 20%
• Shave off roughly 1:15 min from the CI runtime

Metric	Before	After
Total # of factories	157,949	103,761
Total time spent creating factories	17:59.851	11:14.739
Total time	34:41.482	27:03.658

Here is the CI run with all the changes merged together:

[lago-claude-ai-agent]

Automated pre-review (advisory, not a required check) — verdict: HOLD · CI green

HOLD — several new let_it_be fixtures are mutated by examples, so later examples can reuse stale in-memory Active Record state even though the database transaction rolls back.

• Keep membership per-example or use reload/refind in both membership specs; revoke changes its status, while update examples cache differing role associations.
• Apply the same isolation to the shared add_on/plan records that are discarded in the one-off, subscription-creation, and subscription-amendment execution specs.