Fix max_position_embeddings handling in NomicBertConfig and FlashNomicBertConfig

[alvarobartt]

What does this PR do?

This PR fixes the handling for max_position_embeddings in both NomicBertConfig and FlashNomicBertConfig so that instead of using max_trained_positions as the default value, which was only required when the model was flagged with custom code on the Hugging Face Hub, since now those config.json have been updated to be natively supported without custom code as per e.g., https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/commit/e9b6763023c676ca8431644204f50c2b100d9aab.

Additionally, this PR also solves an issue with using alias for n_positions for the default value of max_position_embeddings when it was not provided in the config.json, to instead not use alias (as when the keys in config.json are duplicate it will raise an error) and use max_trained_positions as the fallback, assuming that both max_position_embeddings and max_trained_positions are optional, but at least one needs to be set, in that order of priority.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline?
• Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the documentation guidelines.
• Did you write any new necessary tests? If applicable, did you include or update the insta snapshots?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.