fix(anthropic): normalize ParsedTextBlock to avoid Pydantic serialization warnings

[Zelys-DFKH]

Description

Anthropic SDK >= 0.84.0 returns `ParsedTextBlock` objects in `content_block_start` events. `ParsedTextBlock` is a `TextBlock` subclass with an extra `parsed_output` field that is not part of the discriminated union on `Message.content`. Calling `model_dump()` on those events triggers `PydanticSerializationUnexpectedValue` warnings on every agent invocation (6+ warnings per call), cluttering stderr even though serialization succeeds and results are correct.

The fix adds a `_normalize_event()` static method that converts `ParsedTextBlock` back to a plain `TextBlock` before `model_dump()` is called. This preserves all content data (text, type, citations) and removes the spurious `parsed_output` field that causes the discriminated-union mismatch.

Key details:

• The import of `ParsedTextBlock` is guarded with `try/except ImportError` for SDK versions that do not ship it
• Non-`content_block_start` events pass through unchanged
• Events with a plain `TextBlock` also pass through unchanged

Related Issues

Fixes #1865

Documentation PR

N/A

Type of Change

Bug fix

Testing

• Added three unit tests for `_normalize_event()` covering: ParsedTextBlock conversion, plain TextBlock passthrough, and non-`content_block_start` passthrough
• All 45 existing tests pass

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.