AnthropicChatGenerator does not get correct prompt_tokens when using streaming.

[mathislucka]

Describe the bug
Anthropic returns streaming events where the first chunk has the input_tokens and the last chunk has the completion_tokens.

However, in this line we try to get usage only from the last chunk:

haystack-core-integrations/integrations/anthropic/src/haystack_integrations/components/generators/anthropic/chat/chat_generator.py

Line 416 in d804dd7

last_chunk_meta = chunks[-1].meta

We need to change the conversion so that we get input tokens from the first chunk and output tokens from the last chunk.

To Reproduce
Use the AnthropicChatGenerator with streaming and look at ChatMessage.meta["usage"].

Describe your environment (please complete the following information):

• OS: [e.g. iOS]
• Haystack version:
• Integration version: