Bugfix/257 speed up simpathsmodelpopulatetaxdbreferences process

[andrewbaxter439]

What

• changes donortaxunit lookups at the start of each new run to lazy join policies then sort in memory

Why

• this becomes a very slow process for multiple years of tax benefit policies. Should speed up from ~5 mins to ~2 mins in post-setup pre-run phase

[Copilot]

Pull Request Overview

This PR improves the performance of the SimplPathsModelPopulateTaxdbReferences process by optimizing how donor tax unit lookups are performed at the start of each run. The changes convert eager fetching to lazy loading and perform sorting in memory instead of in the database.

• Changed from eager to lazy fetching of policies in DonorTaxUnitPolicy
• Modified query to remove LEFT JOIN FETCH and sort in memory by max year policy income
• Added database indexes and batch size optimization for better performance

Reviewed Changes

Copilot reviewed 4 out of 5 changed files in this pull request and generated 2 comments.

File	Description
src/main/java/simpaths/model/taxes/DonorTaxUnitPolicy.java	Changed fetch strategy from EAGER to LAZY, added BatchSize annotation and database indexes for performance
src/main/java/simpaths/model/SimPathsModel.java	Modified donor pool query to remove JOIN FETCH and implemented in-memory sorting by max year policy income
src/test/java/simpaths/integrationtest/expected/Statistics21.csv	Updated expected test results with minor numerical differences due to performance optimization changes
src/test/java/simpaths/integrationtest/expected/HealthStatistics1.csv	Updated expected test results with minor numerical differences due to performance optimization changes

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[andrewbaxter439]

To sense-check: this produces miniscule differences in calculated stats, usually beyond the 7th/8th decimal places in only about a 10th of estimates. The only way I think this is possible is due to differing de-duplication procedures in the searching approaches?