chore: remove dead catch_warnings blocks from add_groups migration

[williambdean]

The 'fit_data not defined in InferenceData scheme' warning was emitted by arviz's add_groups, which was replaced by idata["/fit_data"] = fit_data in the PyMC 6 migration (#2616). DataTree assignment never raises this warning, making the catch_warnings / filterwarnings blocks dead code.

Removed in: mixed_logit.py, mnl_logit.py, nested_logit.py, maxdiff.py.

Flagged by @daimon-pymclabs in #2616 review.

---

📚 Documentation preview 📚: https://pymc-marketing--2652.org.readthedocs.build/en/2652/

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.07%. Comparing base (1e2b2ed) to head (bf04485).

Additional details and impacted files

@@            Coverage Diff             @@
##           v1.0.0    #2652      +/-   ##
==========================================
- Coverage   94.07%   94.07%   -0.01%     
==========================================
  Files          97       97              
  Lines       14617    14606      -11     
==========================================
- Hits        13751    13740      -11     
  Misses        866      866              

☔ 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.

[williambdean]

Please take a look @daimon-pymclabs

[daimon-pymclabs]

Took a look — this is good to merge. Checked the mechanical side carefully:

• The four customer_choice fit() methods and the two conftest fixtures all suppressed the same UserWarning ("The group fit_data is not defined in the InferenceData scheme"), which no longer fires now that fit_data is a recognized group — so the blocks are genuinely dead.
• The import warnings removals are correct: dropped from mixed_logit.py, mnl_logit.py, nested_logit.py and both conftests where it's now unused, and retained in maxdiff.py (still used by warnings.warn(...) at ~L601). No dangling references left in any changed file.
• Full test matrix + customer_choice notebooks are green, so the suppressed path is exercised.

One optional belt-and-suspenders: a quick run of the customer_choice fit tests under -W error::UserWarning would prove the warning is gone rather than merely silent, but I don't think it's necessary to block on. LGTM 👍