ci(bench): ERD, charts/groupings and multi-tab SQL scenarios in the app benchmark #1173

Problem

benchmark/app_perf_bench.dart covers grid, tree, sidebar, tabs and typing (#987), but not the newer visual parts, and CI (.github/workflows/benchmark.yml) runs only grid_perf_bench.dart in scroll mode. ERD, Charts and Groupings have no frame budget check, so regressions there go unnoticed.

Scope

Scenarios in the app bench, and a CI job that runs them (informational, like the grid bench; see #1132 for noise):

• ERD on a generated schema (50 / 200 tables): open Diagram, hover across cards, drag a card for 2 s, Auto layout.
• Charts and Groupings on 5 000 and 100 000 rows: switch to Charts, change type / X / Y, change grouping column and aggregate.
• SQL workspace with 5 tabs holding results: run a query in one tab (pane rebuild count, frame times).

Report build / raster p50 / max like the existing scenarios; thresholds per scenario.

Acceptance

• Scenarios run in CI and appear in the bench report.