fix(pipeline): REST fallback for GraphQL batch-level failures (502/timeout)

[rpothin]

Summary

• fall back to per-repo REST hydration when an entire GraphQL batch request fails
• keep recording REST errors per repository so PAT-policy failures are still classified downstream
• add regression coverage for successful REST fallback and REST fallback failure cases

Root cause

Large Microsoft org repositories can make the batched POST /graphql request fail at the transport level (for example 502 Bad Gateway or timeouts). The previous outer catch recorded that single batch error for every repo in the 20-repo batch, so all 20 became detailFailures and blocked the pipeline.

Fix

When the whole GraphQL batch throws, batchGetRepositoryDetails now falls back to getRepositoryDetails() for each repo in the batch. That isolates failures to the repos that truly fail under REST while allowing public repos to continue through the pipeline.

Validation

• cd Pipeline && npm test
• cd Pipeline && npm run typecheck