Add remote client to langchain_ollama

[antoninoLorenzo]

Checked

• I searched existing ideas and did not find a similar one
• I added a very descriptive title
• I've clearly described the feature request and motivation for it

Feature request

In the Ollama classes in langchain_ollama there is no possibility to set a remote client.
For example, in OllamaEmbeddings: instead of using ollama.embedding(...) there should be a ollama.Client configured; the default endpoint will still be localhost, hower for remote Ollama inference one will do something like OllamaEmbeddings(model='...', host='...').

Motivation

This functionality would make the Ollama wrappers more flexible. This would improve the library experience for open source solutions.

Proposal (If applicable)

No response