fix(pdf): surface the underlying error when a PDF fails to load

[jsgrrchg]

When a PDF fails to load, the viewer runs the raw error through classifyPdfError and only shows a friendly message. Anything that doesn't match its keywords becomes "An unexpected error occurred while loading this PDF.", and the original error is never logged. That is exactly what the reporter of #519 sees, and right now neither the user nor we can tell what actually failed.

This also hid one of our own messages: "This PDF can no longer be previewed because it is outside the active vault." has no matching keyword, so it was shown as the generic error too.

Changes:

• The error view gains a Show Details toggle that reveals the raw pdf.js error (monospace, selectable) with a Copy button. The toggle only appears when the friendly message hides information.
• Every failure is logged via logError("pdf", "Failed to load PDF", { path, error }), so it ends up in the renderer log even if the user never opens the details.
• The "outside the active vault" message is now shown as-is.
• The error view is extracted into PdfErrorView, and the duplicated button styles into PdfErrorButton.

This doesn't fix the root cause in #519; it's meant to capture the real error so we can diagnose it. For reference, the PDF attached to that issue loads and renders without errors locally with pdf.js 6.2.108 using the viewer's exact options (including stopAtErrors: true), both in Node and in headless Chrome, canvas and text layer included.

Related to #519.

Validation:

• PdfTabView tests: 15 passed, including two new ones (details toggle and copy; outside-vault message without details).
• tsc --noEmit -p tsconfig.app.json and eslint on the changed files passed.
• Not exercised manually in the running app.