Avoid scanning entire Geodatabase file

It appears an entire scan of a Geodatabase file is needed before a single record can be returned. The following pulls one record from an 870 MB file and takes ~36 seconds to do so.

```sql
select FIPS_CODE FROM st_read('SF_Professional_AK.gdb/a00000011.gdbtable') LIMIT 1;
```

```
┌───────────┐
│ FIPS_CODE │
│ varchar │
├───────────┤
│ 11111 │
└───────────┘
```

```
Run Time (s): real 36.234 user 33.853635 sys 2.386451
```

Start by reproducing the SQL query against SF_Professional_AK.gdb/a00000011.gdbtable and trace the st_read entry point in the C extension. Compare the work performed for LIMIT 1 with the reported 36-second full-file scan; done means a single record can be returned without scanning the entire 870 MB file.

Written by the indexing model from the issue text.