Support get_columns and parse protobuf object for Python SDK

Here is the code to test.

```
import sqlalchemy as db
from sqlalchemy import inspect

zk_cluster = "127.0.0.1:2181"
zk_path = "/openmldb"

engine = db.create_engine(f'openmldb:///db1?zk={zk_cluster}&zkPath={zk_path}')
connection = engine.connect()

insp = inspect(engine)
print(insp.get_table_names())

columns = insp.get_columns('employees')

for column in columns:
print(column)
```

We may add the code in `openmldb_dialect.py`.

```
def get_columns(self, connection, table_name, schema=None, **kw):
"""
columns_info = [
{
'name': 'name',
'type': sqlalchemy.String(),
'nullable': False,
'default': None,
'primary_key': False,
'autoincrement': False,
'comment': 'Unique identifier for the row'
}
]
return columns_info
"""
raise NotImplementedError()
```

The issue is about implementing get_columns and protobuf parsing in openmldb_dialect.py. Start by examining the existing dialect code to understand the connection and table structure. You'll need to fetch column metadata from OpenMLDB, likely via its API or SQL, and map types to SQLAlchemy types. Check if there are existing tests for the dialect to run your implementation against.

Written by the indexing model from the issue text.