feat: Unfreeze Document in Haystack 2.0

[vblagoje]

As a part of the ongoing improvements for #5945 let's unfreeze Document as it is blocking for #5923

[anakin87]

The implementation itself is ok.

I'm just thinking about the consequences of unfreezing this class.
Some thoughts expressed by @ZanSara:

• Re-evaluate frozen Documents #5899
• feat: OpenAIDocumentEmbedder #5822 (comment)

[vblagoje]

I say we unfreeze and then follow the recommendations @ZanSara outlined in #5899 @masci you were confident about unfreezing this morning, but we didn't elaborate. Are we ok to unfreeze?

[masci]

I don't get why the ID should be kept in sync with the content, or even the embeddings of a Document. If users provide an ID themselves, that's not on Haystack to decide if the ID is valid or not; if users don't provide an ID, Haystack will generate one for them trying to minimize conflicts, but that's it. All the major vector databases allow to update embeddings for the same ID, why Haystack shouldn't?
@vblagoje curious why you suggest we should follow the recommendations in #5899?

[vblagoje]

@masci mostly because I didn't participate in that conversation so I suggested it as a middle ground 😀