fix: warehouse graceful termination and avoid marshalling proc error job payloads

[deepshekhardas]

This PR combines fixes from two key PRs:

PR rudderlabs#6767 - Warehouse Graceful Termination:

• Replace panic with proper error handling in warehouse router
• Files changed: warehouse/router/identities.go, warehouse/router/state.go

PR rudderlabs#6872 - Avoid Marshalling Proc Error Job Payloads:

• Introduce procErrorJob type to avoid expensive marshal/unmarshal round-trip
• Update processor to use new type for error job handling
• Files changed: processor/pipeline_worker.go, processor/processor.go

Changes:

• 4 files changed, 22 insertions(+), 12 deletions(-)
• Added procErrorJob struct wrapping jobsdb.JobT with parsed events
• Removed unnecessary jobsdb import from pipeline_worker.go
• Improved error handling in warehouse graceful termination

---

Summary by cubic

Gracefully handles warehouse router termination by replacing panics with proper errors, and reduces processor overhead by avoiding marshal/unmarshal for error jobs. Prevents crashes and cuts CPU/GC in error paths.

• Bug Fixes

  • Replaced panics with error returns and logs in warehouse/router (manager init, schema marshal, upload ID scan); state.go now returns an empty state on invalid input instead of panicking.

• Performance

  • Introduced procErrorJob wrapping *jobsdb.JobT with parsed events to avoid marshal/unmarshal; updated maps to map[string][]procErrorJob in processor and pipeline_worker. Removed an unused jobsdb import.

^{Written for commit 0d05b4e. Summary will update on new commits.}

[deepshekhardas]

Closing this combined PR. Will create separate PRs for rudderlabs#6767 and rudderlabs#6872 for easier review.