I finally followed the following method for my requirement:

1) Connected the data verse tables from PowerApps to synapse workspace using the synapse link

2) Used the standard template provided in it to convert the multiple .csv files to parquet format

3) Translated the missing columns in the converted parquet file using the metadata information as given here in Access Dataverse choices

It is still difficult to translate the metadata within databricks since there could be multi-level relationships. But replicating the PowerBI dataverse connector seems to be currently not possible. So I am able to temporarily resolve this issue.