Some backends are subclassed and others are monkey patched - better to do it one way

**Is your feature request related to a problem? Please describe.**
This is code refactoring, clarity and improvement

**What version of DVT are you using?**
8.7

**What type of connections are you using for source and target?**
Postgres

**Describe the solution you'd like**
DVT uses backend in two different ways
1. The Oracle, MSSQL model - where the a Backend child class with BaseAlchemyBackend as the base class
2. The Postgres, MySQL model - where the ibis backend is monkey patched to add libraries.

Both models provide similar functionality, customized do_connect, list_primary_keys, raw_metadata, is_char_padded etc. The first appears cleaner, the second is difficult to understand.

Better if we go with one model, perhaps the first - thoughts?

Read the Oracle and MSSQL backend subclasses alongside the Postgres and MySQL implementations that monkey patch ibis. Compare how do_connect, list_primary_keys, raw_metadata, and is_char_padded are provided, then define which single backend model should replace both approaches while preserving that functionality.

Written by the indexing model from the issue text.