test: add source-side integration test for crawl/parse pipeline

[cloudshiftchris]

Summary

• Adds MavenSyncIntegrationTest — a Kotest FunSpec that points a real MavenHttpRepository at a user-supplied source repo and runs everything the engine does (crawl, directory-listing parsing, maven-metadata.xml parsing, target-diff, per-version listArtifactVersionAssets) up to but excluding any upload to the target. The target is faked with an empty in-memory stand-in so every source version appears "missing".
• The test is gated on MAVEN_SYNC_IT_CONFIG (path to a JSON file using the standard SyncConfig schema, layered over defaults.json via the same Hoplite pipeline as MavenSyncMain); when unset the test is reported as skipped and ./gradlew test does no network IO.
• Writes a deterministic JSON discovery report to build/reports/maven-sync-it/discovery-report.json (group:artifact → version → asset filenames), asserts ≥1 artifact / ≥1 version / ≥1 asset per version, and supports regression snapshots via MAVEN_SYNC_IT_EXPECTED.
• README gains an "Integration tests" section with env vars, example config, run command, and the note that target.url is required by the schema but not contacted.

Test plan

• ./gradlew test — passes, integration test reports SKIPPED.
• MAVEN_SYNC_IT_CONFIG=<local.json> ./gradlew test --tests '*MavenSyncIntegrationTest*' against a real source repo.
• Snapshot mode: rerun with MAVEN_SYNC_IT_EXPECTED pointing at a saved report.

🤖 Generated with Claude Code