fix(linear): prevent unrelated team cache failures from blocking batch push

[cursor]

Summary

• fix Tracker.BatchPush so non-primary team state-cache failures are recorded per-team instead of aborting the entire push
• preserve strict behavior for primary team cache failures (still fail fast for create path)
• stop falling back to primary team cache for routed updates when a team cache is unavailable, and emit issue-scoped errors instead

Bug and impact

In multi-team Linear setups, BatchPush eagerly built state caches for every configured team and returned a top-level error on the first failure. A permission or availability issue in any secondary team prevented all pushes, including healthy-team updates.

Root cause

Global preflight cache initialization treated all team cache failures as fatal for the whole batch run, rather than scoping failure to the team that could not be resolved.

Validation

• added regression test: TestBatchPush_NonPrimaryStateCacheFailureDoesNotAbort
• ran:
  • CGO_ENABLED=1 go test -tags gms_pure_go ./internal/linear -run '^TestBatchPush_NonPrimaryStateCacheFailureDoesNotAbort$'
  • CGO_ENABLED=1 go test -tags gms_pure_go ./internal/linear