feat(mistralai): remove tenacity retries for embeddings

[noeliecarbo]

MistralAI (and embeddings in particular) is the only package that uses tenacity retries.

This PR removes them, offering the following benefits:

• consistency with other packages
• users can more easily customize retries according to their needs.

[codspeed-hq]

CodSpeed Performance Report

Merging #33491 will not alter performance

_{Comparing noeliecarbo:noeliecarbo/feat_remove_tenacity_retries_from_mistralai_embeddings (7457d95) with master (6e25e18)}

Summary

✅ 1 untouched
⏩ 33 skipped¹

Footnotes

1. 33 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[eyurtsev]

@noeliecarbo thanks for the contribution!

This is a breaking change, but it's not clear to me what users gain from it. What kind of customization would they want to do that's meaningful in practice from what they can already do?

[noeliecarbo]

@noeliecarbo thanks for the contribution!

This is a breaking change, but it's not clear to me what users gain from it. What kind of customization would they want to do that's meaningful in practice from what they can already do?

@eyurtsev Thanks for the feedback! From my perspective, I see the following benefits:

• For my use case, I need to catch specific HTTP 400 errors (too many inputs). It seems that the official MAX_TOKENS of the Mistral API is inaccurate and varies somewhat inexplicably.
• Users may need different retry behaviors than those provided by retry_if_exception_type and wait_fixed (tenacity provides many more options).

[eyurtsev]

what happens if you set max_retries = 0?