docs: document enable_computational_runtime on agent settings

[Joetib]

Syncs the skills with iblai/iblai-dm-pro#3002 (part of iblai/iblai-platform#2412).

What changed upstream

enable_computational_runtime is now readable. The flag lives on the agent rather than on its settings row, so it was accepted by the settings PUT but never came back on the GET — a client could set it and then not be able to read it back.

Doc changes

iblai-api-agent-setting only:

• added to the settings PUT body field list;
• a note under Notes that it is now returned by the GET as well, having previously been write-only.

Scope note — the lookup change was reverted

This PR originally also documented agent lookups as unique_id-only, matching an upstream change that made a slug or name return 404.

That upstream change has been reverted (iblai/iblai-dm-pro@57a5f8a61): production orgs still address agents by slug, so enforcing unique_id would have broken live traffic across ~25 endpoints. The REST endpoints continue to accept unique_id, slug or name exactly as before, so those doc edits are gone from this PR:

• iblai-api-agent-session/SKILL.md is back to its original wording and no longer appears in the diff at all.
• iblai-api-agent-setting/SKILL.md keeps only the settings-field addition; the "only the unique id resolves" note is removed.

{mentor} in the path-vars line still reads "the agent's unique id" — that is the pre-existing wording, and it remains the identifier you should prefer, it is just not the only one the API accepts.

Branch name is stale. It is still sync-mentor-unique-id-lookups from the original scope. Kept rather than reopened so the cross-references from the upstream PR stay valid; the commit and diff reflect the real, narrower scope.