Currently, if httpfs extension has not been loaded yet:

COPY (SELECT 42 AS answer) TO 's3://my_bucket/my_file.parquet';

either fails like:

IO Error:
Cannot open file "s3://my_bucket/my_file.parquet": No such file or directory

where s3://my_bucket/my_file.parquet is treated as a local file system and can't find the base directory s3://my_bucket that is necessary to actually save the file.

LOAD httpfs;
COPY (SELECT 42 AS answer) TO 's3://my_bucket/my_file.parquet';

will rightly fail like:

HTTP Error:
Unable to connect to URL https://my_bucket.s3.us-east-1.amazonaws.com/my_file.parquet?uploads=: Forbidden (HTTP code 403)

that, given I choose a random bucket name that I have no permissions for is appropriate.

Goal of this PR is making autoloading (or detection that loading an extension is needed) work in more broader contest, in particular:

• for targets of COPY TO
• for globs (this worked already, but now with the same infrastructure)
• for fetching single files in a non-globbing situation (example: SELECT name, id FROM ICEBERG_SCAN('https://raw.githubusercontent.com/duckdb/duckdb-iceberg/main/data/persistent/equality_deletes/warehouse/mydb/mytable', VERSION='1', allow_moved_paths = true); will now correctly trigger autoloading of httpfs)

This infrastructure is also easy to generalize, it's a matter of calling VirtualFileSystem::FindFileSystem(path) overload that takes either a ClientContext or a FileOpener: VirtualFileSystem::FindFileSystem(path, context_or_file_opener).

After this PR, same example as above would either autoload and throw HTTP Error:

HTTP Error:
Unable to connect to URL https://my_bucket.s3.us-east-1.amazonaws.com/my_file.parquet?uploads=: Forbidden (HTTP code 403)

or throw a Missing Extension Error:

Missing Extension Error:
File s3://my_bucket/my_file.parquet requires the extension httpfs to be loaded

Please try installing and loading the httpfs extension by running:
INSTALL httpfs;
LOAD httpfs;

Alternatively, consider enabling auto-install and auto-load by running:
SET autoinstall_known_extensions=1;
SET autoload_known_extensions=1;

Fixes #19180
Fixes autoloading of FileExist, example:

SELECT name, id FROM ICEBERG_SCAN('https://raw.githubusercontent.com/duckdb/duckdb-iceberg/main/data/persistent/equality_deletes/warehouse/mydb/mytable', VERSION='2', allow_moved_paths = true);