Default to a sparse 64 x 48 world at 5% occupancy

[ppericard]

Change

Pierre asked to make the sparse world used in previews and studies since #8/#9 the launch default:

	Before	After
Grid	32 × 24 (768 cells)	64 × 48 (3,072 cells)
Initial occupancy	30%	5%
Initial agents	230 (58/58/57/57)	153 (39/38/38/38)

Only ExperimentConfig::default() changes. Preset mix, maintenance, seed, 500-tick duration, pacing and all model rules are unchanged. --width, --height and --occupancy still override the default.

Tests

Several Rust and browser tests set their own dimensions but inherited the former 30% occupancy. With the new default, an 8 × 6 world would drop from 14 to 2 agents, which would silently weaken them. These tests now pin occupancy: 300_000 / --occupancy 0.3, so their populations, the fsm-seed1-tick10 fixture and every assertion stay as they were. Tests that use the complete default configuration now exercise the new default.

ordinary_launch_uses_the_mixed_fsm_population now also checks the default dimensions, occupancy, 153 initial agents and 39/38/38/38 groups. It fails on main with left: (32, 24, 300000) right: (64, 48, 50000).

Docs

README and docs/development.md describe the new default. The two running examples drop their now-redundant --width 64 --height 48 --occupancy 0.05 flags. The population_trace study command keeps explicit dimensions for reproducibility. The unreferenced files in docs/screenshots/ still show the older world and were not regenerated.

Verification (local Windows, Rust 1.95.0)

• cargo fmt --check, Clippy in both feature modes with -D warnings: passed
• cargo test without default features, with web, and in release: 77 / 95 / 77 passed
• npm run check: 7 passed
• git diff --check: clean
• headless --ticks 0 reports width=64 height=48 occupancy=0.050000 initial_count=153
• Not run locally: Playwright e2e (Chromium isn't installed on this machine), cargo audit, npm audit. These are left to the Linux and Windows CI jobs for this commit.

🤖 Generated with Claude Code