fix(dataset): normalize tokenizer config and change hash from tokenizer class to tokenizer path

[NanoCode012]

Fix hash across tokenizers. We should not base on tokenizer class name as they could have different vocabs.

Description

Motivation and Context

How has this been tested?

Screenshots (if appropriate)

Types of changes

Social Handles (Optional)

[NanoCode012]

Test failing due to not normalizing config, I believe.