Demote non-actionable report_error! calls to log levels (drive + server)

[warp-dev-github-integration]

What & why

Audit of every report_error! call in the Warp Drive and cloud-sync files listed in the request, to reduce Sentry noise by demoting calls that don't represent actionable engineering issues. Guided by .agents/skills/logging-and-error-reporting/SKILL.md.

Only logging severity changes — no logic or control-flow changes. Demoted calls become log::warn! (a breadcrumb, not a Sentry issue); error detail is preserved via {err:#}.

Rubric used

• Demote when the failure is caused by factors outside client-code correctness: network/RPC failures, disk/IO errors, invalid user/enterprise config, cloud-object lifecycle races, malformed data received from the server, and by-design capacity limits. These are already "not actionable" — an engineer can't act on them, and several are auto-suppressed to Warn by report_error!'s is_actionable() anyway.
• Keep report_error! for violations of internal invariants that signal a client bug: programming guards (often paired with debug_assert!), state-machine violations, and persistence-channel failures (a broken writer channel can mean real data loss).

Demoted → log::warn! (15)

Drive:

• drive/sharing/dialog/mod.rs (×3) — WeakViewHandle<TerminalView> upgrade failures when removing/updating/inviting session guests. Expected lifecycle race (the terminal view was dropped); already handled by early return, and the same file uses warn! for analogous handled cases.

Server:

• server/telemetry/mod.rs — "More telemetry events in queue than the limit to persist": by-design capacity condition (we intentionally keep only the last N).
• server/telemetry/secret_redaction.rs — failed to build the telemetry secrets regex: driven by invalid user/enterprise-configured patterns; falls back to the prior regex.
• server/telemetry/collector.rs (×3) — telemetry disk-write failure (IO), shutdown flush failure (network; sibling timeout arm already warn!), and persisted-file removal failure (IO).
• server/server_api.rs (×2) — notify_login POST send failure (network) and access-token retrieval failure; both best-effort in a fire-and-forget path.
• server/server_api/ai.rs — skipping a single malformed ambient-agent task while deserializing a list (handled skip; the code comment already says "log the error and skip").
• server/server_api/object.rs — per-object GSO conversion failure (handled skip; previously interpolated the object uid into the grouping message, fragmenting Sentry).
• server/cloud_objects/listener.rs — websocket connection failure after retries (network; matches the sibling retry-pending arm).
• server/cloud_objects/update_manager.rs (×2) — unparseable task_id received from the server (malformed server data) and single-object fetch failure (network).

Kept as report_error! (actionable — documented for the reviewer)

• drive/panel.rs (×6) — owner can't be resolved for a user-initiated create: a user action silently no-ops; worth investigating.
• drive/index.rs (×6) — current workspace missing from the dropdown list (state inconsistency) and four "wrong action / not-yet-supported" programming guards.
• drive/workflows/modal.rs, workflow_arg_type_helpers.rs, import/{modal_body,nodes,queue}.rs — internal invariants ("both/neither id specified", "state should be path expanded", "folder id should exist", "counter should exist").
• server/server_api/ai.rs (×3, unchanged) — unknown GraphQL enum variants ("update client GraphQL types"). These are the skill's own canonical report_error! example, are actionable (client types are stale), and are already throttled with ReportErrorLogMode::OncePerRun.
• server/cloud_objects/update_manager.rs — persistence-channel send failure + data-consistency/programming guards (debug_assert!-paired).
• server/sync_queue.rs (×4, unchanged) — dependency-resolution invariants that the queue's own dependency tracking should prevent.
• server/telemetry/context.rs — serializing a fixed struct should never fail.
• cloud_object/model/persistence.rs (×4) and cloud_object/mod.rs — persistence-channel sends + a state-machine invariant.

Validation

• ./script/format — passes.
• CARGO_BUILD_JOBS=1 cargo clippy -p warp -- -D warnings — passes (edits are confined to non-test source in the warp crate; unused report_error imports were removed where the last use was demoted).

Conversation: https://staging.warp.dev/conversation/7786c619-c2b7-4a6a-a410-68f473e9d7bb
Run: https://oz.staging.warp.dev/runs/019f6cfe-7a4f-7b40-97b2-68c4511d1d85

This PR was generated with Oz.