fix(runtime-host): log why a transcript page read failed

[MoonOld]

Summary

SessionContinuityCoordinator#readTranscriptPage collapses every error that is not a TranscriptPageRequestError into persistence_failed / "Session transcript is unavailable", and returns it as an operation outcome. Because the handler returns instead of throwing, the [runtime-host] unexpected <operation> failure log in operation-dispatcher.ts — the only caller of boundedFailureDiagnostic — never runs for this operation. The cause therefore reaches neither the Runtime Host log, the desktop diagnostic report, nor the Renderer, which falls back to generic copy ("任务内容暂时无法刷新，请稍后重试。").

This records the cause with boundedFailureDiagnostic before the generic outcome is returned. The outcome the caller receives is unchanged; the only new output is one [runtime-host] log line.

Refs #5572

Verification

• npm --workspace @maka/runtime-host run typecheck — clean
• npx biome check on both touched files — clean
• node --test packages/runtime-host/dist/__tests__/session-continuity-coordinator.test.js — 47/47 pass
• Fails without the change: with the source hunk reverted and the test kept, the new case fails with AssertionError (0 pass / 1 fail)

The new test injects a reader that fails only on the page path — bootstrap requests carry no position, page requests always do — then asserts the generic outcome is still returned and that console.error was called exactly once with the injected cause:

✔ a failed transcript page records the underlying cause before the generic outcome
ℹ tests 47 / pass 47 / fail 0

Not run: the full workspace test suite.

Review focus

Whether logging directly here is right, versus routing it through onPublicationFailure. The sibling bootstrap path (session-continuity-coordinator.ts:966) calls onPublicationFailure under a comment saying the failure "has to leave a trace here", but that hook is wired to context.requestDrain (execution-composition.ts:925), not a logger — so no trace is emitted there either. That path is deliberately untouched here; it is context for #5572, not part of this change.

AI use

Select exactly one:

• No generative tool made a substantive contribution
• Generative tooling made a substantive contribution

Tool(s) and scope: Maka (a DeepSeek-backed agent running in Maka Desktop) drafted the log line and the regression test, and ran the verification commands above. The root-cause analysis that motivated the change was reviewed and corrected before this PR by the author.

Checklist

• Tests cover the change and fail without it
• Lint, format, typecheck and the affected suites pass locally

Does this PR entail a change in behavior?

• Yes — described under Summary above
• No