Bump clone multipliers to hit ~40M/~100M dataset targets

[jathavaan]

Problem

Synthesized dataset sizes fall ~17% short of targets set in #255. Last setup run (release 2026-05-16.1):

size	target	actual	multiplier (incl. originals)
small	~5M	4,166,773	1× (passthrough)
medium	~40M	33,334,184	8× (7 clones + original)
large	~100M	83,335,460	20× (19 clones + original)

Root cause: targets in DatasetSize.clones_per_polygon were sized assuming a ~5M base, but the conflated source dataset is 4.17M polygons.

Fix

Bump clones_per_polygon in src/domain/enums/dataset_size.py:10:

• MEDIUM: 7 → 9 clones → ~41.7M rows
• LARGE: 19 → 23 clones → ~100.0M rows

Update docstring targets in the same enum.

Validation

Re-run setup container, confirm buildings_medium ≈ 40M and buildings_large ≈ 100M (minus dropped invalid clones).

Refs #255.