Record existing automatic naming performance

[marcinpsk]

Parent

#73

What to build

Add a manually invoked test-suite performance runner and use it to record the existing automatic naming implementation before production family behavior changes. The runner must exercise real NetBox models, committed callbacks, and PostgreSQL. It must produce a normalized baseline that the completed refactor can run again on the same hardware.

Acceptance criteria

• The runner covers no matching rule, one plain rename, four-channel structural creation, rename of an installed parent with four channels, deferred channel reconciliation, and virtual-chassis reapplication with one and eight families.
• Complete model-save signal paths and direct deferred callbacks are measured as distinct scenarios.
• A PostgreSQL work-profile pass records normalized statements, execution plans, selected indexes, planner cost, actual rows and loops, buffer access, temporary blocks, rows affected, WAL activity, and scaling behavior.
• PostgreSQL profiling observes the real statements without replaying mutating SQL, and node timing is disabled.
• A separate uninstrumented pass records warm-up behavior, raw wall-time samples, process CPU samples, median, and p95.
• Results record source, NetBox, Python, and PostgreSQL revisions together with planner settings and fixture sizes.
• SQL parameter values and environment-specific data are excluded from artifacts.
• The runner uses an isolated PostgreSQL test database and fails when required profiling support is unavailable.
• The runner remains outside default test discovery and is not added to CI.
• The existing-implementation baseline and a readable summary are retained as implementation evidence.
• No production naming behavior changes in this ticket.

Blocked by

None — can start immediately.

[marcinpsk]

Implemented in PR #83: #83

The retained baseline uses NetBox feature revision d7f79bd50162b2b8a8d07b52348d332bd9404f3c, PostgreSQL 18.3, Python 3.14.4, 15 samples, and 3 warmups across 14 scenarios. Complete model-save paths and direct deferred callbacks are recorded separately.

Validation completed:

• performance runner passed all 14 scenarios
• full NetBox feature suite passed 680 tests with 12 skipped and experimental query-count update mode
• Ruff check and format checks passed
• artifact schema, safety, and callback-isolation audits passed
• standards and issue-spec reviews have no remaining findings

The PR uses Closes #74, so merging it will close this ticket.