fix(profile): resolve no-scheme URLs and ignore placeholder URN drift

[peggyrayzis]

Closes

Closes #20

Description

Fixes two li profile regressions in production: no-scheme profile URLs were parsed as raw usernames, and recipient cache drift warnings could surface synthetic placeholder URNs.

Summary

• Normalize no-scheme LinkedIn URLs in parseLinkedInUrl (for example linkedin.com/in/<handle>).
• Tighten LinkedIn hostname validation to linkedin.com or *.linkedin.com only.
• Filter recipient cache warning comparisons to canonical URNs (urn:li:member:<digits> or urn:li:fsd_profile:ACo...) so placeholder cache values do not emit false drift warnings.
• Resolve recipient-cache path lazily from env to make tests/processes isolate cache files safely.
• Add regression coverage for no-scheme URLs, malicious suffix-match hostnames, no-scheme recipient resolution, and warning behavior with placeholder vs valid cached URNs.
• Isolate Vitest recipient cache path to avoid polluting user-level temp cache between test and production runs.
• Append a memory-log entry for the regression/fix.

Checks

• npm run check
• npm run security
• bash scripts/check-docs.sh

Review

• Reviewer findings addressed or documented
• Residual risks documented

Residual Risks

• Canonical URN filtering intentionally treats non-ACo... fsd_profile IDs as non-cacheable; if LinkedIn introduces a new profile-ID shape, warnings for that shape will be suppressed until patterns are updated.
• Integration tests for this path remain skipped unless auth-backed test env is enabled.