Add configurable max_batch_size for GoogleVertexAI embedding service settings

[dan-rubinstein]

Issue - #137288

This change adds a configurable max_batch_size to the GoogleVertexAI embedding service settings. This will allow users to configure the number of chunks during chunked inference (ingest to a semantic text field) to the downstream service. This will allow users to override the existing 250 batch size to either reduce it to avoid hitting the 20k token limit across all chunks sent in a single request or to increase it (up to the 250 max) for improved inference performance if they are not sending large enough chunks to hit the limit.

[elasticsearchmachine]

Hi @dan-rubinstein, I've created a changelog YAML for you.

[elasticsearchmachine]

Pinging @elastic/ml-core (Team:ML)