[Security solution] Create new class instance for each use of LangChain chat model

[stephmilovic]

Summary

Follow up from #189810
Related to #189771

We need to create a new class instance for each use of any LangChain chat model. This is necessary to avoid any potential side effects from shared state and a variety of bugs.

1. Renames getLlmInstance to createLlmInstance
2. Adds a comment to explain to always create a new instance of the LLM class and to never assign it to a variable.
3. Removes anywhere llmClass was assigned to a variable (llm) and instead always calls createLlmInstance()

Tests

ESQLKnowledgeBaseTool	AlertsCountTool	OpenAndAcknowledgedAlertsTool	Alert Summary
Azure 4o	Azure 4o	Azure 4o	Azure 4o
Gemini Pro 1.5	Gemini Pro 1.5	Gemini Pro 1.5	Gemini Pro 1.5
Sonnet 3.5	Sonnet 3.5	Sonnet 3.5	Sonnet 3.5

[elasticmachine]

Pinging @elastic/security-solution (Team: SecuritySolution)

[kibana-ci]

💚 Build Succeeded

• Buildkite Build
• Commit: 569975a

Metrics [docs]

✅ unchanged

To update your PR or re-run it, just comment with:
@elasticmachine merge upstream

[andrew-goldstein]

Thanks for this fix @stephmilovic! 🙏
✅ Desk tested locally
LGTM

[stephmilovic]

Another example, this will solve the titling step from invoking tools: https://smith.langchain.com/public/e34f8013-12cc-49c2-b68a-b42205322c9c/r

[kibanamachine]

💚 All backports created successfully

Status	Branch	Result
✅	8.15

Note: Successful backport PRs will be merged automatically after passing CI.

Questions ?

Please refer to the Backport tool documentation

[mistic]

This PR didn't make it on time of the latest v8.15 BC. Updating the labels.