[fix][evaluation] write NULL (not 0) for absent evaluator score to fix N/A miscounting in aggregation

[xueyizheng]

Background

A nil/null score is a legitimate signal in the evaluator protocol for expressing "Not Applicable (N/A)": when an evaluator cannot judge an input (e.g. an image with no text), it returns a nil score plus a reasoning. score is optional in the SPI DTO and a pointer in the DO by design. Such records can carry status=Success:
Async report path: status is supplied by the external agent callback (POST /v1/loop/evaluation/evaluators/result); the converter does not require a score when status=SUCCESS, and ReportEvaluatorInvokeResult performs no status/score consistency check.
Sync prompt path: a successful parse yields Success, and a nil score is not rejected.
Problem

UpdateEvaluatorRecordResult (async path, introduced in #421) used a value-typed var score float64, so a nil score was coerced to 0 and persisted as 0 (not NULL).

Previously this was harmless: aggregation deserialized output_data and skipped nil scores (hasScore=false), never reading the score column. After #590 switched aggregation to read the score column directly (to avoid OOM), the filter status=Success AND score IS NOT NULL can no longer exclude a literal 0, so N/A records are miscounted as 0 and drag down the average / distribution.

Fix
Change the score parameter from float64 to *float64 in both the repo impl and the DAO, writing NULL (not 0) when there is no score, so the score column semantics are honest.
This aligns the async update path with the sync create path (ConvertEvaluatorRecordDO2PO, which has always written a *float64 / NULL).
Add DAO-level tests for NULL vs. concrete-value writes, a repo-level nil case, and regenerate the mock signature.
After this fix, the narrow aggregation query's score IS NOT NULL filter is fully consistent with the old in-memory "skip nil" semantics.
━━━━━━━━━━━━━━━━━━━━━━━━━

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

@@           Coverage Diff           @@
##             main     #592   +/-   ##
=======================================
  Coverage        ?   77.51%           
=======================================
  Files           ?      679           
  Lines           ?    80914           
  Branches        ?        0           
=======================================
  Hits            ?    62720           
  Misses          ?    14463           
  Partials        ?     3731           

Flag	Coverage Δ
unittests	77.51% <100.00%> (?)

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

Files with missing lines	Coverage Δ
...tion/infra/repo/evaluator/evaluator_record_impl.go	91.01% <100.00%> (ø)

---

Continue to review full report in Codecov by Harness.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6d2dc02...d1948de. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.