chore(enterprise): bump the submodule pointer to main (8331df5 -> b5080b6)

[dolho]

Summary

Bumps src/backend/enterprise from 8331df5 to b5080b6, current enterprise main. The one new commit is abilityai/trinity-enterprise#818: seats (#811), per-person seats (#814), drift on read and three-state consent (#817).

That is the private half of #3271. #3271's public seam (assignment_provider.resolve_seat / resolve_served_seat) reads the provider's optional seat_for / served_seat_for. The adapter at 8331df5 implements neither, so on dev every seat holder's Workspace role card and objective join read no_seat. That is #3534, which keeps rows A1 and A6 of the 1.0 testing matrix (abilityai/trinity-enterprise#847) red.

Checks

• Migration graph (docs/ENTERPRISE.md, "Before advancing the recorded submodule pointer"): check_alembic_heads.py → 21 revisions, 1 head (0020_seat_holders), PASS.
• Enterprise assignments suites against public dev (84ad179fd) with the CI env: test_500 ×3, test_606, test_811, test_814, test_818 → 195 passed.
• Live, local stack on dev + enterprise b5080b6: boot logs Trinity Enterprise modules registered, no registration failure. Agent with no x-role, a canon role file, an objective owner: role:project-lead and a primary seat for the viewer → GET /api/enterprise/client-portal/agents/{name}/role shows role Project Lead, seat_source: person, both owned objectives, the number (7 / 10, behind). GET /api/agents/{name}/objectives no longer reports no_seat. On 8331df5 the same setup gives no_seat and no objectives.
• Staged diff is the gitlink only (8331df5..b5080b6); b5080b6 is on enterprise main.

Fixes #3534

🤖 Generated with Claude Code

[obasilakis]

Pointer is a one-commit fast-forward to enterprise main (enterprise#818, approved there). Migration graph at b5080b6: 21 revisions, 1 head. CI green.