[Fleet] Prevent installation of packages containing ML models during preconfiguration/setup

[kpollich]

Fleet cannot guarantee idempotency when installing ML models, so it makes sense for us to disallow this for now.

• Add a step to Fleet setup that excludes any packages that contain at least one ML model asset when installing managed/preconfigured packages
• Ensure we output a log message indicating the package that was excluded and why

[elasticmachine]

Pinging @elastic/fleet (Team:Fleet)

[kpollich]

@joshdover - I was actually just taking a deeper look at this and it seems like we might not actually have an idempotency issue w/ ML model installation:

https:/elastic/kibana/blob/main/x-pack/plugins/fleet/server/services/epm/elasticsearch/ml_model/install.ts#L86-L94

When Fleet installs ML Models, we specifically ignore resource_already_exists_exception exceptions from Elasticsearch. So in a case where multiple Kibana nodes attempt to run setup, I think we'd just silently error on a redundant ML model installation.

[kpollich]

I spoke with @alvarezmelissa87 last week about this since she implemented all of Fleet's support for ML assets in integrations. She had no concerns about idempotency or duplication as a result of multiple concurrent installs of ML assets, so it seems like we can close this out.