Fix silent failures of Alt+O / Alt+C shortcuts

[dream-creator]

Fixes #2634

What this PR does

Alt+O (open original page) and Alt+C (copy original link) currently fail completely silently when the rendered entry has no ._attribution ._attribution-link element, and Alt+C additionally had a leftover console.log and ignored clipboard promise rejections. This PR shows a transient notice when the link is unavailable, surfaces clipboard failures instead of swallowing them, and removes the debug logging.

Investigation notes

While debugging why the shortcuts "do nothing" for some users and work for others:

• The cached entry data on documents.devdocs.io does contain the _attribution block (verified against e.g. git-status.html), so the selector input exists at the data level.
• The failure mode is therefore environment- or page-dependent (stale service-worker caches serving pre-attribution HTML, docs whose scrapers render no attribution, or pages where attribution isn't the last child) — exactly the split reported in the issue (works on some macOS setups, dead on Windows/others).
• Because both handlers bailed out with a bare return, every one of those cases produced identical "nothing happens" behavior, making the bug impossible to self-diagnose from the UI. Surfacing a notice turns the invisible state into actionable feedback.

Changes

• entry_page.js: both handlers now show a transient notice ("The original page link is not available for this documentation") instead of silently returning; onAltC handles the navigator.clipboard.writeText rejection with its own notice; removed stray console.log
• notice_tmpl.js: added noOriginalLinkNotice and copyFailedNotice templates following the existing pattern
• New notices auto-dismiss after 3 seconds and replace each other rather than stacking

Testing

DevDocs has no JS unit-test infrastructure yet (#2673), so verification was: syntax check on both modified files plus manual code-path review against the cached-entry structure linked above. The notice path can be exercised by running any entry whose scraped HTML lacks an attribution block.

Happy to iterate if maintainers would prefer different wording/behavior (e.g. also attempting URL reconstruction for docs without attribution).