Tweak agent builder's llm selection details (#3051)

[Emma
confirmed](https://elastic.slack.com/archives/C056H3WA1BL/p1758189381969239?thread_ts=1757949389.230719&cid=C056H3WA1BL)
that the setting to change the default LLM is _not_ going to be
available before 9.2.0, so we shouldn't reference it in the first docs.

We shouldn't document the non-UI path, as the API to change this setting
is `/internal/`. We'll be OK to communicate this directly to the
customers we invite to the private preview, but otherwise we shouldn't
document this until the setting is available to use in the UI.

Also fixed a typo.

Also took out an explicit reference to Claude 3.7, after confirming with
@joemcelroy and @maxjakob that we don't want to advertise exactly what
the Elastic LLM is using under the hood.

Author: seanstory

solutions/search/agent-builder/limitations-known-issues.md

@@ -16,12 +16,10 @@ These pages are hidden from the docs TOC and have `noindexed` meta headers.
 
 ## Model selection
 
-Initally, {{agent-builder}} only supports working with the [Elastic Managed LLM](kibana://reference/connectors-kibana/elastic-managed-llm.md) running on the Elastic Inference Service which uses Claude Sonnet 3.7, on {{ech}} and {{serverless-full}}. 
+Initially, {{agent-builder}} only supports working with the [Elastic Managed LLM](kibana://reference/connectors-kibana/elastic-managed-llm.md) running on the Elastic Inference Service, on {{ech}} and {{serverless-full}}. 
 
 Locally this picks the first AI connector available.
 
-This can be changed under Stack Management -> AI -> GenAI Settings -> Default AI Connector.
-
 ## Known issues