update new aliasing code, add constants and test cases

[AmeyHengle]

The PR aims to address issues in the current codebase related to intent aliasing and preprocessing steps.
Associated Ticket: https://vernacular-ai.atlassian.net/browse/CMS-2524

Issues:

1. In the current code structure intent aliases are defined as a dictionary mapping {intent: intent_alias}. This differs from the structure commonly used in eevee or production (ref), which makes it difficult to use the alias.yaml files directly in dialogy.
2. Aliasing is present as a preprocessing step in train.py but not in test.py.
3. There is no provision to define independent train and test aliases, which is necessary for multiple clients.
4. Certain preprocessing steps are absent in test.py (make_label_column_uniform, make_data_column_uniform, etc).

Updates:

1. Introduced a separate function for intent aliasing.
2. Provision to define separate train and eval aliases.
3. Check for duplicates and/or malformed values in alias files.
4. Common preprocessing functions in train.py and test.py moved to a single utility file.
5. Added test cases covering the new aliasing code.

[dakshvar22]

@AmeyHengle Can you please describe the problem better? It's not clear what exactly are you solving here.