fix: reduce flakiness in retrieval config integration test

[Hweinstock]

Problem

The integration test test_session_manager_with_retrieval_config_adds_context is flaky (#432). It fails because the relevance score filter (relevance_score=0.7) drops retrieved memories when the semantic similarity between the query "What do I like to eat?" and the extracted preference about sushi scores below 0.7. The test failed 3/3 attempts (original + 2 reruns) in this CI run.

CI logs confirm: Retrieved 1 memories appears but Retrieved X customer context items never does — the relevance filter drops the memory before context injection.

Solution

Two changes to make the test more reliable:

1. Lower relevance_score from 0.7 to 0.3

2. Change query from "What do I like to eat?" to "What sushi do I like?" — semantically closer to the stored preference, producing higher and more consistent relevance scores.

Testing

The test validates the same invariant: retrieval config with LTM causes <user_context> to appear in agent messages. The changes only reduce non-determinism in the relevance score path without weakening the assertion.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (main@d8c91cb). Learn more about missing BASE report.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #437   +/-   ##
=======================================
  Coverage        ?   91.06%           
=======================================
  Files           ?       60           
  Lines           ?     5072           
  Branches        ?      775           
=======================================
  Hits            ?     4619           
  Misses          ?      257           
  Partials        ?      196           

Flag	Coverage Δ
unittests	91.06% <ø> (?)

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.

[notgitika]

LGTM thanks for looking into this!