fix: Update Document.__eq__ to intelligently compare floats

[nigamgauri]

Related Issues

• fixes Update Document.__eq__ to intelligently compare floats #8412

Proposed Changes:

Document.__eq__ performed strict dictionary equality comparison (self.to_dict() == other.to_dict()) on the serialized representation of two documents. This caused equality checks to fail when float fields (such as score, embedding values, sparse_embedding values, or nested floats in meta) differed very slightly due to floating-point imprecision.

This PR replaces the strict comparison in Document.__eq__ with a recursive structure comparison helper (_recursive_is_close) that uses math.isclose (with a relative/absolute tolerance of 1e-7) when encountering floating-point numbers.

How did you test it?

• Added four new unit tests in test/dataclasses/test_document.py verifying float comparisons for score, embedding, sparse_embedding, and nested meta fields.
• Verified that all unit tests, format/lint checks (hatch run fmt), and type checks (hatch run test:types) pass successfully.

[vercel]

Someone is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

gauri nigam seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[sjrl]

Hey @nigamgauri please sign the CLA agreement #11828 (comment) otherwise we can't accept your contribution.

I'll proceed with a review once it's signed.