Commission step can be missing when parsing from Open Data for live texts

[mdamien]

> anpy-cli parse http://www.assemblee-nationale.fr/dyn/15/dossiers/amelioration_tresorerie_associations
[
    {
        "assemblee_id": "15-amelioration_tresorerie_associations",
        "assemblee_legislature": 15,
        "assemblee_slug": "amelioration_tresorerie_associations",
        "beginning": "2018-10-17",
        "long_title": "Société : amélioration de la trésorerie des associations",
        "proposal_type": "PPL",
        "steps": [
            {
                "date": "2018-10-17",
                "id_opendata": "L15-VD187069DI",
                "institution": "assemblee",
                "source_url": "http://www.assemblee-nationale.fr/15/propositions/pion1329.asp",
                "stage": "1ère lecture",
                "step": "depot"
            },
            {
                "date": "2018-11-29",
                "id_opendata": "L15-VD187234S36552",
                "institution": "assemblee",
                "stage": "1ère lecture",
                "step": "hemicycle"
            }
        ],
        "urgence": false,
        "url_dossier_assemblee": "http://www.assemblee-nationale.fr/dyn/15/dossiers/amelioration_tresorerie_associations"
    }
]

Would be a good opportunity to add regression testing to the open data parser too.