EventBrite scraper broken

[Obersand]

Several issues given an EventBrite update sometime in September:

• Element IDs changed. The code crashes looking for div.location-info__address. I got this scanning this Fresque du Plastique event for Switzerland and this Fresque de l'Economie Circulaire event for UK. It looks like the tag became span.start-date-and-location__location. Same for span.date-info__full-datetime which became time.start-date-and-location__date and several others.
• New date format, with the year omitted, an am/pm scheme and strings that depend on a language setting (ex: "pm" vs. "nachm"). I'm actively working on this in sub-issue Erreurs d'analyses de dates EventBrite #75.
• New address format, with the address in the workshop containing just the street (ex: "33 Spring Street"). This causes the geocoder to time out and crash the program (I'll catch that). The full address is sometimes elsewhere on the page, sometimes the city name is in the event title and nowhere else. A longer-term fix would be issue Emission des erreurs de géolocalisation #74 and pushing on the workshop owners to fix this. I may be able to implement a workaround in the meantime.

[thomas-bouvier]

Depending on your decision regarding #80, I believe the Eventbrite scraper is now in good shape. The recent improvements, especially in b05ad41, have significantly enhanced its reliability by switching from Selenium to Playwright.

Playwright uses the browsers' native automation protocols for better reliability, and this resulted in much less flakiness in my experience. Playwright seems ideal for JS-heavy sites like Eventbrite.