Makes inference endpoint the primary way to download and deploy ELSER and E5

[szabosteve]

Overview

This PR overhauls the ELSER and E5 conceptual pages:

• indicates that the recommended and easiest way to download and deploy the models is using inference endpoints,
• introduces the procedure of creating an inference endpoint in the Download and deploy section,
• makes the alternative download and deploy methods less prominent by putting them into collapsible sections that are closed by default,
• propagates the semantic_text workflow for semantic search.

Preview

• ELSER
• E5

[github-actions]

A documentation preview will be available soon.

• 🔨 Buildkite builds
• 📚 HTML diff
• 📙 Preview page

Request a new doc build by commenting

• Rebuild this PR: run docs-build
• Rebuild this PR and all Elastic docs: run docs-build rebuild

_{run docs-build is much faster than run docs-build rebuild. A rebuild should only be needed in rare situations.}

_{If your PR continues to fail for an unknown reason, the doc build pipeline may be broken. Elastic employees can check the pipeline status here.}

[maxhniebergall]

This looks good, but I just want to double check with Jan that adaptive allocations made it into the 8.15 release before I approve this.

[szabosteve]

@maxhniebergall Thanks for the review! I just merged the adaptive allocations-related documentation yesterday. Jan asked me to draft the docs and he approved the PRs. You can find these here:

• API docs
• Conceptual docs
• tracking issue

However, I can change the API requests in this PR to NOT use adaptive_allocations if you prefer.

[szabosteve]

@maxhniebergall I replaced the adaptive_allocations settings with num_allocations and num_threads via e11530e

[maxhniebergall]

Looks great! Thanks for the quick fixes. We will definitely want to re-add the adaptive allocations in 8.16.