fix(hosted): stabilize workspace ownership

[dajiaohuang]

Problem

Hosted Worker used worker:<ephemeral port> as the owner stored in its workspace marker. A retry or process restart on another port therefore could not reclaim the same persisted conversation workspace, even though the trusted Host intentionally selected the same path.

Solution

• require the trusted supervisor to pass a stable, per-workspace --workspace-id
• hash that ID together with the canonical workspace path into an opaque marker owner
• keep the existing marker/path/owner checks, active-workspace protection, TTL/LRU/capacity limits, symlink exclusions, and fail-closed cleanup behavior unchanged
• document that the ID is Host-managed and must never come from player/model input

The path binding prevents one stable ID from silently naming another workspace. A different owner still cannot claim an existing marker.

Integration contract

Supervisors launching sagasmith-agent-worker must add:

--workspace-id <stable-host-issued-workspace-id>

Retries and restarts for the same persisted workspace must reuse the same value even when the worker port changes. Different workspaces must receive different values.

Verification

• uv run ruff check nanobot --select F
• uv run ruff check nanobot/apps/hosted_workspace.py nanobot/apps/hosted_worker.py tests/apps/test_hosted_workspace.py tests/apps/test_hosted_worker.py
• uv run python -m pytest -q tests/apps (26 passed)
• uv run python -m pytest -q tests (5390 passed, 73 skipped)

Regression tests cover a port-changing restart/retry, path binding, cross-owner rejection, ambiguous ID rejection, and the required CLI contract.