Add multimodal support for CodeAgent

[codegen-sh]

Description

This PR adds multimodal support to the CodeAgent, allowing it to process images in Slack messages. This is the second part of the implementation, building on PR #11444 in the cloud repository which adds the ability to capture image information from Slack messages.

Changes

1. Updated the LLM class to support multimodal content:

   • Added a _process_messages_for_multimodal method that converts image URLs in the format [Image: filename](URL) to the appropriate multimodal format for Claude 3 models
   • Modified the _generate method to process messages for multimodal content when using Claude 3 models

2. Updated the CodeAgent class to use multimodal models by default:

   • Added a multimodal parameter (default: True) to the constructor
   • When multimodal is enabled, ensures a Claude 3 model is used even if a non-multimodal model was specified

How to Test

1. Send a message with an image in a Slack thread
2. Mention the CodeAgent bot
3. Verify that the bot can see and process the image in its response

Related PRs

This PR builds on PR #11444 in the cloud repository, which adds the ability to capture image information from Slack messages.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

codegen-bot seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}