Add thread_pool_executor to constructor

[billytrend-cohere]

The PR introduces changes to the Client and AsyncClient classes in src/cohere/client.py.

Summary

The changes add a _executor attribute to both classes and set its value during initialization. The ThreadPoolExecutor is now passed as an argument during initialization, allowing for more flexibility and customization.

Changes

• The _executor attribute is added to the Client and AsyncClient classes.
• During initialization, the thread_pool_executor argument is added to the __init__ method of both classes.
• The value of self._executor is set to the provided thread_pool_executor argument.
• The previous instantiation of _executor outside the methods is removed.

[fede-kamel]

Hi! I wanted to follow up on this PR as I've just submitted PR #699 which complements this implementation.

This PR addressed part of issue #534 by allowing ThreadPoolExecutor configuration at the client level, which is great. However, issue #534 also mentioned the need for configurable batch_size, and some users might need per-call flexibility rather than client-wide settings.

PR #699 adds:

• batch_size parameter to the embed() method (the other key request from Allow users to configure embed_batch_size or ThreadPoolExecutor size when calling Client.embed #534)
• max_workers parameter for per-call thread pool override
• Dynamic optimization without recreating the client

Both PRs work together nicely - users can set a default ThreadPoolExecutor via your implementation and then override it for specific operations using the per-call parameters.

Just wanted to connect these implementations since they're solving related parts of the same issue!