Use search_name as fallback for TIGER imports when update tables are dropped

[AyushDharDubey]

Closes #2463

This PR adds a fallback lookup that uses the persistent search_name table when the update information tables have been dropped. The existing lookup logic remains the primary path; only when an undefined_table exception is raised do we resolve the road via search_name instead.

The new getNearestNamedRoadPlaceIdSlow function leverages the existing idx_search_name_name_vector GIN index, which efficiently filters candidates by checking overlap between name_vector and the provided token_info, keeping the fallback reasonably fast (still slower than getNearestNamedRoadPlaceId).

[lonvia]

Thanks, that's the general direction I had in mind. I added a few inline comments,

There is a little gottcha here: the table search_name might not exist either, when Nominatim was set up in reverse-only mode. You could in that case fall back to doing name matching on placex entries or just tell the user that import is not possible.

[AyushDharDubey]

Hello, thanks for the reviews. I’m still getting familiar with parts of the codebase. Going through the sql-preprocessor code was very helpful.

the table search_name might not exist either, when Nominatim was set up in reverse-only mode. You could in that case fall back to doing name matching on placex entries or just tell the user that import is not possible.

based on my understanding, falling back to name matching on placex would be prohibitively slow for practically large databases, thus I’ve leaned towards explicitly reporting that the import is not possible in this configuration.

[lonvia]

I've tested this against a planet database. It's fairly fast but it leaves quite a few interpolations unmatched because the street name doesn't fit well enough. It would be good to have a replacement for getNearestParallelRoadFeture() as well which uses the placex table (again address_rank between 26 and 27 only) to find any parallel road.

[AyushDharDubey]

Thanks for testing over planet db. I have added placex lookups for both parallel and nearest roads respectively, also squashed and force-pushed for cleaner commits.