Use Unaccent with dandiset search filter

[jjnesbitt]

Closes #1941

This change normalized accent characters when searching dandisets. For example (from that issue), the words Buzsaki and Buzsáki would now resolve to the same word in search.

It seems the introduction of this change has no real impact on the query performance. In fact, perhaps due to now using a less complex search function overall (since we're no longer using the DRF SearchFilter class), the performance seems to be ever so slightly better.

Postgres docs for the unaccent extension: https://www.postgresql.org/docs/current/unaccent.html
Django docs for unaccent: https://docs.djangoproject.com/en/5.1/ref/contrib/postgres/lookups/#unaccent

Of note, we can't simply use the __unaccent__ lookup without any other changes, as is shown in the Django docs, because the metadata field is a JSONField, and that lookup only works on Charfield and Textfield.

[kabilar]

Thanks @jjnesbitt. cc @bendichter

[jjnesbitt]

Just realized I never wrote any tests for this. I'll do that.

[waxlamp]

Instead of the noqa directive here, why not just follow the recommendation to eliminate the unnecessary assignment?

[jjnesbitt]

This function is parsing the query param, and performing a number of actions on it. In this case, it's just that one .replace call, but realistically, you could see us adding to it in the future. It just felt a bit awkward/arbitrary to just return param.replace('\x00', '') 🤷‍♂️ .

[waxlamp]

I think just go with ruff's recommendation here. "We might change the code in the future" doesn't seem compelling enough to override its style advice.