fix: use str instead of float to preserve precision

[1yam]

This PR addresses an issue where using float to represent certain numerical values resulted in loss of precision due to rounding. By switching to str, we can maintain the exact value without rounding errors.

[github-actions]

Summary:
The PR involves a small change where the required_tokens type from float to str is updated in the PriceResponse class. This change is unlikely to introduce bugs and is primarily focused on improving data consistency and type safety.

Highlight:

-    required_tokens: float
+    required_tokens: str

This change is straightforward and does not affect the overall functionality of the codebase, making it suitable for a routine review.

[1yam]

Unit test should work need PR merge

[hoh]

Converted to Draft, this should use the Decimal type in Python and ensuring that the upstream API returns the expected type.

[Psycojoker]

Apparently the other PR has been merged, should we resume this PR?