A few things:

1. I think having the line number output in tests is going to cause a ton of diff noise when changing files.

2. I think it's worth doing the more difficult thing to integrate into the normal postgres errors:

   The alternative would be to reproduce what elog(ERROR, ...) does and expose some of the elog.c internals.

   Afaict all that's needed is to inline the ereport/elog implementation, similarly to what we already do in include/pgduckdb/logger.hpp. And then change the call to errfinish to take the correct file, line number & function name.

3. Changing InfokeCPPFunc like is a great start, but it's only half of the issue. I think we'd also want to to change PostgresFunctionGuard to store the original errdata in the exception, and propagate it upwards. Otherwise you'll always get the line info from the final wrapper location, not from the actual source. This could be done in a future PR though.