Python: preserve native keys in filtered vector delete tools

[Sunhy (Shy7777)]

Motivation & Context

A filtered delete tool fails on MongoDB collections with ObjectId keys, even though deleting the same key without a filter works. The filtered path converts retrieved keys to JSON strings and passes them straight to the connector, which expects native keys.

Description & Review Guide

• What are the major changes? Convert matched keys back through the collection's existing key_from_json hook before deletion. Add a MongoDB regression covering a matching subset, a mapped tenant field, and an empty match.
• What is the impact of these changes? Filtered deletion works with native key types while processed_keys remains JSON-compatible. Public APIs, approval settings, and filtering behavior are unchanged. The matching regression fails on the base revision; the fix passes 113 MongoDB unit tests and 385 core vector/in-memory tests. Changed-file lint, formatting, and source type checking also pass. Live database integration tests were not run.
• What do you want reviewers to focus on? Whether the native-key conversion belongs at this point between filtered retrieval and deletion, and whether the regression covers the connector interaction adequately.

Related Issue

Fixes #8534

No other open PR was found for this issue.

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
  • All 498 related unit tests pass locally; the full repository matrix remains for CI.
• The PR follows the Contribution Guidelines
• This PR is linked to an issue and there is no other open PR for this issue (see Related Issue above).
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix) — a workflow keeps the label and title prefix in sync automatically.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.