web: bound monitoring and distinguish SDR execution from pipeline backlog

[hyunmoon]

Draft — depends on #1533. Files changed includes the foundation; incremental review shows only this PR's changes.

Summary

Bound monitoring work and distinguish executing SDR attempts from queued and remaining pipeline sectors.

• Provide bounded Cluster Tasks snapshots and PoRep pages with batched enrichment, explicit stale/unknown states and compatible RPC fields. Took starts at recorded current Do entry, not posting or ownership. Pause and late responses cannot manufacture fresh elapsed time.
• Count distinct sectors in one server-side summary query. The miner table and Overview share a snapshot/request, with running, preparing, queued, unlinked and exceptional SDR categories. Remaining workload is not labeled in-flight.
• Post-SDR counts exclude failed/complete sectors but include outstanding Finalize/MoveStorage after Commit confirmation.

Defaults remain pending preview 500 and coalescing enabled, without sealing-specific priority or default pending hiding. This requires the parent's attempt schema/entry recording, not pacing or backend-health policy.

Validation and limits

API normal/race, exact-tag vet/lint, frontend tests and offline Chromium validation. PostgreSQL query/handler fixtures cover transitions, missing/old/wrong tasks and query failure. Large-backlog fixtures include tens of thousands of waiting sectors linked to actual SDR task rows, plus multiple owners and running/preparing tasks. The summary query/mapper was also exercised on disposable Yugabyte 2025.2.2.2-b11 with the actual curio-schema migration runner.

These single-node cost samples are not production or multi-node performance proof; attributable retries were not measured. Execution-entry telemetry does not prove healthy native computation.

Review scope

Review boundary	Commits	Diffstat
Own changes, after foundation	5	67 files changed, 6284 insertions(+), 219 deletions(-)
Cumulative against upstream main	8	113 files changed, 12073 insertions(+), 635 deletions(-)

Series overview. No dependency on the other children. Ready after the foundation is accepted, its duplicated diff is removed and applicable checks pass.

Own commits and comparison base

Own commits:

• 39ed163f752f377e90d1891970075a3964d320f4 — web: show bounded current-attempt task snapshots
• 13c36379deab17796e17e6846850c03e669db2e9 — web: bound PoRep page queries and recover refresh state
• f7cb60769b4138035035d0bd7fc0a7ace98e38c3 — web: distinguish SDR attempts from remaining pipeline sectors
• 245b2c02e82667d0edc0f5e2fb58856be5d8a935 — test: label overview fixture population and browser version
• eba504dfeeee0f07c370fc877dec6dce548887e7 — test: exercise large linked SDR summary populations

Cumulative counts use upstream 86e95967cd602dbf8db5ea538abe986c3968e2e3; the incremental base is 0c1f7cbd4dc9797e9773f1edd165af6b72d3fe27.