fix: make three-period ITS test fixtures order-independent

[anevolbap]

Summary

Fixes #959.

datetime_data and integer_data in test_three_period_its.py drew from the session-scoped rng fixture, so the generated data depended on how many random draws earlier tests consumed. Any PR that changed test collection (like #851, which removed one parametrize case in an unrelated file) shifted the stream and could flip test_analyze_persistence_sklearn, whose persistence_ratio >= 0 assertion only holds for some draws.

Changes

• Both fixtures now use their own seeded generator instead of the shared session rng.
• Dropped the persistence_ratio >= 0 assertions in the pymc and sklearn persistence tests. A negative ratio is a legitimate outcome when the counterfactual sits above the observed post-period.

Testing

• The reproduction from test_analyze_persistence_sklearn is order-dependent through the session-scoped rng fixture #959 (pytest causalpy/tests/test_instrumental_variable.py causalpy/tests/test_three_period_its.py) now passes: 54 passed.
• Pre-commit hooks pass on the commit.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.16%. Comparing base (984163e) to head (c9d7d35).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #960   +/-   ##
=======================================
  Coverage   95.16%   95.16%           
=======================================
  Files          93       93           
  Lines       15030    15030           
  Branches      896      896           
=======================================
  Hits        14304    14304           
  Misses        507      507           
  Partials      219      219           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.