make sure output has articles first, then redirects for each article

[mtmail]

So far the output was sorted by title only, so the city of Athens, Greece (Q1524) printed

zgrep ^en wikimedia_importance.tsv.gz | grep Q1524$
en    r    Ahtens    0.7868465180438662    Q1524
en    r    Aktiki    0.7868465180438662    Q1524
en    r    Atenás    0.7868465180438662    Q1524
en    r    Athenae    0.7868465180438662    Q1524
en    r    Athenai    0.7868465180438662    Q1524
en    r    Athēnai    0.7868465180438662    Q1524
en    r    Athence    0.7868465180438662    Q1524
en    r    Athenes    0.7868465180438662    Q1524
en    r    Athénes    0.7868465180438662    Q1524
en    r    Athènes    0.7868465180438662    Q1524
en    r    Athenian    0.7868465180438662    Q1524
en    r    Athenians    0.7868465180438662    Q1524
en    a    Athens    0.7868465180438662    Q1524
en    r    Athens,_Attica    0.7868465180438662    Q1524
en    r    Athens_Basin    0.7868465180438662    Q1524
en    r    Athens_city_center    0.7868465180438662    Q1524

Nominatim's tools/refresh.py imports only one (the first) title per wikidata_id.

In this case that would be a common typo, there was never an article for Ahtens in wikipedia. Earlier this year it was Aktiki, a historical name from hundreds of years ago.

We want to make sure the article comes first:

en	a	Athens	0.7868465180438662	Q1524
en	r	Ahtens	0.7868465180438662	Q1524
en	r	Aktiki	0.7868465180438662	Q1524
en	r	Atenás	0.7868465180438662	Q1524
en	r	Athenae	0.7868465180438662	Q1524
en	r	Athenai	0.7868465180438662	Q1524
en	r	Athēnai	0.7868465180438662	Q1524
en	r	Athence	0.7868465180438662	Q1524

I would've added more logic into Nominatim's import logic (`tools/refresh.py) but Nominatim doesn't require the type column.

        The file must be a gzipped CSV and have the following columns:
        language, title, importance, wikidata_id

        Other columns may be present but will be ignored.