Add AMBAR music dataset to Lenskit

[sushobhan2024]

issue #1142

• Added amber.py to load data from its path
• Updated the describe.py so imported dataset can be described with the existing describe command

[mdekstrand]

A few things this needs off the bat:

• tests (tagged with @mark.realdata, see the Amazon and Steam load tests for examples). The tests should probably look at it in data/ambar.
• adding to data/convert.py.
• adding to the data conversion CLI test script.

[codecov]

Codecov Report

❌ Patch coverage is 2.43902% with 40 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.17%. Comparing base (928f7eb) to head (d9cc10c).

Files with missing lines	Patch %	Lines
src/lenskit/data/sources/ambar.py	0.00%	36 Missing ⚠️
src/lenskit/cli/data/describe.py	20.00%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1187      +/-   ##
==========================================
- Coverage   90.37%   90.17%   -0.21%     
==========================================
  Files         261      262       +1     
  Lines       17376    17417      +41     
==========================================
+ Hits        15704    15705       +1     
- Misses       1672     1712      +40     

☔ 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.

[mdekstrand]

Now that the tests are there, they are failing: https://jynx.mde.one/repos/1/pipeline/54/3

The test environment has a copy of the AMBAR folder from the data set repo at data/ambar.

[sushobhan2024]

The error is due to duplicate user_ids. There are 8 duplicate user_ids.
5 have conflicts (contradictory country/continent, one even contradictory gender). The other 3 are repeated records.
I am considering removing those 5 records, and keeping first of the 3 repeated records.

Also, there are 8 different user_ids in ratings_info.csv that have no row at all in users_info.csv. Removing these as well, and passing missing="error" instead of missing="insert" in add_interactions().

conflicting user_ids (from users_info duplicates): 5
{506586384, 39168978, 128783640, 390604169, 38606714}

truly missing user_ids (in ratings, absent from users_info): 8
{271316318, 306287460, 163187558, 839347986, 871284436, 955505332, 183305018, 387251006}

overlap between the two groups: 0

ratings from conflicting users: 1018
ratings from truly missing users: 747
total distinct user_ids being excluded: 13

1,765 of 3,311,462 total ratings (0.053%) from these 13 users are now excluded.

Let me know if excluding is the right approach here.

[mdekstrand]

Probably better to keep the ratings with missing users, those users just won't have attributes. On the conflicts, dropping is probably fine, we could also keep them and just pick the first or last record (if the records have timestamps, we should pick the last one by timestamp).

[mdekstrand]

@mdekstrand just FYI, for security reasons I have to manually approve the woodpecker tests every time they run. They can't be re-triggered from within a push.

[sushobhan2024]

This dataset does not have timestamps.