Add skill endorsement operations

[KirillSergeevich]

Summary

Adds the SDK surface for the st.endorsePersonSkill action and the st.doForSkills iteration in Linked API.

• endorsePersonSkill.execute({ personUrl, skillName }) endorses one skill on a 1st-degree connection's profile by its exact name; letter case and extra spaces are ignored. The result is void. Failures arrive in the usual errors list: personNotFound, connectionNotFound, skillNotFound, endorsementUnavailable, limitExceeded.
• endorsePersonSkills.execute({ personUrl, skillNamePatterns? }) endorses every skill whose name matches any of the case-insensitive RE2 patterns, or every skill when skillNamePatterns is omitted. The workflow is st.openPersonPage → st.retrievePersonSkills → st.doForSkills → st.endorsePersonSkill.
  • Data: Array<{ skillName, success, error? }>. It lists only the selected skills, in profile order.
  • Per-skill errors stay on that skill's item, for example limitExceeded on the rest of the list after the daily cap, or connectionNotFound.
  • Workflow-level errors give no data and arrive in errors. These are failures to open the profile or read its skills: personNotFound, selfProfileNotAllowed, or a profile-view limitExceeded.
• Adds skillNotFound, endorsementUnavailable and limitExceeded to the action error vocabulary. stEndorsements is already in TLimitCategory on main (Add stEndorsements limit category #11).
• Version 2.3.19.

Rebased on main 540df6e. The branch keeps main's formatting of the files #11 touched.

Verification

• npm run typecheck, npm run lint and npm run build pass.
• npm test: 3 suites, 82 tests. That includes a new src/operations/__tests__/endorse-operations.test.ts, which exercises both operations through their public execute/result path with the HTTP layer mocked. It covers:
  • the exact request JSON, with and without patterns;
  • success;
  • every error type;
  • mixed per-skill results, with unselected skills dropped;
  • the capped tail;
  • root and read failures;
  • an empty selection.
• npm pack --dry-run ships the new operations and no compiled tests.