Data Type Bug Fixes

[cgivre]

This PR addresses multiple bugs and code quality issues across the sqlalchemy-drill codebase.

Bug Fixes

• URL encoding issue: Fixed malformed USE queries when Superset URL-encodes schema paths (e.g., %2F for /). Added urllib.parse.unquote() to properly decode database paths before
  processing.
• Data type detection: Fixed issue where all column types were returning as UserDefinedType:
  • Expanded _type_map with additional Drill types: varbinary, numeric, float, float4, real, tinyint, char, character, string, null, struct, array
  • Fixed extraction of type names from DBAPITypeObject.values instead of converting the object to string
  • Improved precision stripping to handle all formats (e.g., varchar(100), decimal(10,2))
• Splunk schema support: Added Splunk to the list of plugins with dynamic schemas that require querying data directly to retrieve column metadata (similar to MongoDB handling)
• Streaming results option: Added stream_results connection parameter to allow disabling HTTP streaming, which can work around certain Drill REST API bugs with array columns
• Removed obsolete imports: Removed references to DefaultCompiler which no longer exists in modern SQLAlchemy versions