Support LinkedIn Jobs search panel capture

[JoseMelNet]

This PR adds safe support for capturing the active LinkedIn job from the Jobs Search Panel while preserving the existing /jobs/view flow.

Main changes:

• Adds LinkedIn job helpers to detect job context and resolve stable job IDs.
• Canonicalizes all LinkedIn vacancy links to https://www.linkedin.com/jobs/view/<job_id>.
• Uses the canonical vacancy identity for side panel state, history, tasks, and async analysis.
• Adds Search Panel support without changing backend/schema.
• Prevents saving ambiguous Search Panel vacancies when no stable job_id can be resolved.
• Adds stale-vacancy protection before saving.
• Restores and reuses the legacy p/span longest-text description extraction strategy for both /jobs/view and Search Panel, scoped to the active Search container.
• Normalizes extracted descriptions by removing LinkedIn UI prefixes/suffixes such as “Acerca del empleo” and “… más”.
• Adds node:test coverage for LinkedIn helpers.

Backend/schema unchanged.

Validation:

• node --test chrome-extension\tests\linkedin-job-helpers.test.js
• node --check chrome-extension\linkedin-job-helpers.js
• node --check chrome-extension\content.js
• node --check chrome-extension\background.js
• node --check chrome-extension\sidepanel.js
• ..venv\Scripts\python.exe -m unittest discover -s tests -q
• ..venv\Scripts\python.exe -m ruff check .
• ..venv\Scripts\python.exe -m compileall app tests
• git diff --check

Manual validation by user:

• Capturing from /jobs/view/<job_id> works and extracts the real vacancy description.
• Capturing from /jobs/search/?currentJobId=<job_id> works and extracts the active vacancy description.
• Clicking a different job in Search updates the data loaded by the extension.
• The final saved link is canonicalized to /jobs/view/<job_id>.
• The implementation intentionally prioritizes reliable plain-text extraction over exact LinkedIn visual formatting.

Important design note:
A prior attempt to preserve richer Search Panel formatting was removed because it introduced a separate parsing path. The final implementation intentionally reuses the legacy p/span longest-text extraction strategy for both /jobs/view and Search Panel, scoped to the active Search container, prioritizing reliability over visual fidelity.