[antonsoo]

A valid group with 101 tables currently raises ValueError: there seem to be cyclic dependencies between the tables, even when no table has a foreign key:

from csvw import TableGroup
from csvw.db import Database

tg = TableGroup.fromvalue({'tables': [
    {'url': f't{i}', 'tableSchema': {'columns': [{'name': 'ID'}]}}
    for i in range(101)
]})
print(len(Database(tg).tables))

The ordering loop removes one eligible table per pass but stops after 100 passes. Continue while tables remain, and stop as cyclic only when a complete pass makes no progress. Add forward/reversed 103-table dependency-chain round trips and a real two-table-cycle check.

Validation: full suites pass on Python 3.12 (842 passed, 16 expected failures) and 3.9 (845 passed, 13 expected failures), with 100% configured coverage; changed source/test Flake8 passes. Published Concepticon list collections of 101 and 113 tables now import successfully in both table orders, with every stored row and cross-list concept frequency matching raw TSV references. Six complete imports at 2/97/100 tables are unchanged. Source on both Python profiles and a fresh installed wheel agree. Four nonrectangular released list files are explicitly excluded without editing their data.