feat(Lungo-UI): Centralized error handling

[misi-bp]

Overhaul Lungo frontend networking and failure handling: replace axios with a shared fetch stack, centralize URLs and request targets, report API errors consistently, and surface actionable messages across chat, streaming, agentic graph, and modals. Follow-up commits on this branch align reportRequestError labels with fetch targets, fix graph OASF/identity edge cases, and document the pipeline for maintainers.

---

Description

HTTP client and URL layer

• Replace axios with httpFetch and helpers: fetchJson, fetchNdjsonStream, fetchSse (shared SSE frame parsing).
• Introduce HttpError, parseHttpError, and endpointLabel on errors for logging and UI metadata.
• Centralize routes in urls.ts (apiPaths, env-backed bases) and httpRequestTargets.ts / HttpRequestTarget builders; migrate agenticWorkflowsClient and catalog consumers to fetch.
• workflowChatRouting resolves prompt/stream/suggested-prompt URLs from WorkflowSummary.chat_api_target and streaming flags.

Error reporting and global UI

• reportRequestError: structured dev/prod logging; optional userMessage → global notification banner via errorNotificationStore.
• reportUiError, ErrorBoundary with recoverable fallback, and error notification banner on RootPage.
• Logs and banner source prefer HttpError.endpointLabel when the transport layer set it.

Chat and streaming

• Agent prompt (JSON and NDJSON), auction / group / recruiter streaming stores unified on fetchNdjsonStream + reportRequestError; shared ndjsonStreamingStatus / stream teardown hardening.
• Pattern reference documentation and pattern chat use agentic-workflows request builders; errors reported at hook boundaries (no duplicate logging in stream UI callbacks).
• Suggested prompts, transport config, and catalog/bootstrap/SSE failures surfaced with user-visible messages instead of silent failures.
• useApp split into focused hooks (useAppPromptHandlers, streaming effects, workflow catalog, pattern reference, etc.).

Graph, modals, and identity

• Unified modal error presentation (LoadingErrorState, compact negative empty states) for badge, policy, OASF, and about flows.
• reportRequestError in graph modals uses the same endpointLabel as the fetch (badgeDetailsRequest, oasfRecordRequest, pattern chat/documentation builders).
• useNodeTransportInterfaces: pass workflow chat_api_target into fetchOasfRecord (fixes logistics OASF 404s to exchange port); cache empty transport results to avoid refetch loops.
• Directory node: dedicated OASF modal content (directory URL/version); no identity-apps badge/policy actions; safe endpoint labels in identity modal catch when slug resolution fails.
• Directory information modal field typography aligned with Info modal (body2 label/value rows).

UI polish and dependencies

• Conditional TransportRail padding on CustomNode only; dark-mode resting rail label contrast; compact-group layout padding split.
• Content-Type: application/json on JSON POST bodies (fixes 422 on /agent/prompt after axios removal).
• Bump @open-ui-kit/core to 2.2.3; prefer OUK imports over direct MUI where touched; security-related dependency updates and tsconfig adjustments.

Documentation

• Add http-and-error-handling.md — URL builders, transport helpers, reportRequestError contract, and per-feature error presentation map.

Tests

• Coverage for parseHttpError, reportRequestError, ErrorBoundary, workflow chat routing, agentic workflows API parsing, topology/identity UI map, and related regressions.

---

Issue link

Fixes #648

Type of change

• Bugfix
• New feature
• Breaking change
• Refactor
• Documentation
• Other (please describe)

Checklist

• I have read the contributing guidelines
• Existing issues have been referenced (where applicable)
• I have verified this change is not present in other open pull requests
• Functionality is documented (http-and-error-handling.md)
• All code style checks pass
• New code contribution is covered by automated tests
• All new and existing tests pass

[delthazor]

The solution is comprehensive, and very solid. I think it is good overall, I just had small nitpicks.

[pregnor]

LGTM at a glance, looked good on the live show

[misi-bp]

[pregnor]

LGTM still, just fix the CI issues

[delthazor]

Pretty robust, let's see it in action