feat(data): wire clrd2025 into load_sample, docs, and tests (#745)

[SaguaroDev]

Contributes to #745. Stacked on top of @kennethshsu's #745 branch (which adds chainladder/utils/data/clrd2025.csv). Targets casact:#745, not main.

This PR covers steps 2, 3, and 4 from my comment on the issue (the maintainer offered to take #1 and did):

1. load_sample wiring — adds a clrd2025 branch in chainladder/utils/utility_functions.py mirroring the existing clrd config (origin=AccidentYear, development=DevelopmentYear, index=[GRNAME, LOB]). The default columns list uses the modernized CAS Schedule P names: IncurredLosses, CumPaidLoss, BulkLoss, EarnedPremDIR, EarnedPremCeded, EarnedPremNet. Per @kennethshsu's note, PostedReserves2007 is kept in the CSV but not in the default columns list (parallel to how PostedReserve97 is not in clrd's default columns).
2. Docstring — clrd2025 added to the complete-dataset line in the load_sample docstring.
3. User guide — clrd2025 row added to docs/library/sample_data.md.
4. Test — test_load_sample_clrd2025 in chainladder/utils/tests/test_utilities.py asserts:
   • all six LOBs are present (comauto, medmal, othliab, ppauto, prodliab, wkcomp)
   • the modern column names load (IncurredLosses, not IncurLoss)
   • origin starts at 1998 and includes 2007

The existing generic test_load_sample already discovers any new CSV in chainladder/utils/data/, so clrd2025.csv is also covered there.

One behavior to flag for review

The DevelopmentYear column in clrd2025.csv runs through CY 2016 (because AY 2007 + dev lag 10 = CY 2016). The Triangle constructor pads origin to span the full valuation range, so cl.load_sample("clrd2025").shape is (768, 6, 19, 19) — a sparse square with real data in the top-left 10×10 block and NaN elsewhere. Operations like Development().fit(...).ldf_ still produce the right link ratios (verified locally on wkcomp.CumPaidLoss.sum()), but it's not the clean 10×10 a user might expect by analogy with clrd.

Two options if you'd rather see a tight 10×10:

# Option A: trim inside load_sample before returning
tri = tri[(tri.origin >= "1998") & (tri.origin <= "2007")]
tri = tri[tri.development <= 120]

That gives (768, 6, 10, 10) with origin = 1998..2007 and development = 12..120, exactly matching clrd's layout.

I left the trim out of this PR because it felt presumptuous — the current Triangle behavior is consistent and the padding may be intentional. Happy to add the trim in a follow-up commit on this branch if you want it. Let me know which you prefer.

Test run

chainladder/utils/tests/test_utilities.py ..........................   24 passed

CC @henrydingliu — thanks again for the offer to help test.

---

Note

Low Risk
Low risk: changes are limited to wiring a new sample dataset into load_sample plus associated docs/tests, without affecting core algorithms or existing dataset behavior.

Overview
Adds a new clrd2025 built-in sample dataset to cl.load_sample, including dataset-specific origin/development/index settings and updated CAS Schedule P column names (e.g., IncurredLosses).

Updates documentation to list clrd2025 among available samples and adds a focused unit test asserting expected LOBs, value columns, and accident-year range for the new dataset.

^{Reviewed by Cursor Bugbot for commit 46103d0. Bugbot is set up for automated code reviews on this repo. Configure here.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (#745@4cf4560). Learn more about missing BASE report.

Additional details and impacted files

@@           Coverage Diff           @@
##             #745     #771   +/-   ##
=======================================
  Coverage        ?   86.08%           
=======================================
  Files           ?       86           
  Lines           ?     4923           
  Branches        ?      638           
=======================================
  Hits            ?     4238           
  Misses          ?      486           
  Partials        ?      199           

Flag	Coverage Δ
unittests	86.08% <100.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 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.

[henrydingliu]

i'm seeing strings like 'incurredlosses' show up three times across three files. can we store the metadata of available sample datasets and fields somewhere, and use it to

• run load_sample()
• key off tests
• generate manifest.in
• generate sample_data.md

[SaguaroDev]

@henrydingliu good point — that triple-duplication isn't unique to clrd2025; the same load_sample config + test + docs-table pattern repeats for every sample in chainladder/utils/data/ (clrd, berqsherm, xyz, the friedland family, etc.). A proper fix would centralize the metadata for all of them and key tests / sample_data.md / MANIFEST.in off of one manifest.

I filed it as a separate issue so this PR can stay scoped to landing the new data: #774. Happy to take a swing at the refactor there once #745 / #771 is in.