fix: Drop await from LDAIClient factory calls

[jsonbailey]

Summary

The LDAIClient factory methods create_model, create_agent, and create_agent_graph were converted from async def to plain def in launchdarkly-server-sdk-ai 1.0.0 — they do no async work, so the await was misleading. The run() methods on the returned ManagedModel / ManagedAgent / ManagedAgentGraph objects remain async.

This PR drops await from the three affected example call-sites in features/, updates the inline doc comments to match, and pins the SDK for those examples at >=1.0.0,<2.0.0.

Requirements

Requires launchdarkly-server-sdk-ai>=1.0.0 (released, on PyPI).

Related SDK PR: launchdarkly/python-server-sdk-ai#187 (feat!: Make AI config factory methods synchronous).

Examples updated

• features/create_model/create_model_example.py — dropped await on aiclient.create_model(...); pin bumped.
• features/create_agent/create_agent_example.py — dropped await on aiclient.create_agent(...); pin bumped.
• features/create_agent_graph/create_agent_graph_example.py — dropped await on aiclient.create_agent_graph(...); pin bumped.

features/create_judge/ and all getting_started/* examples were untouched: create_judge() was already sync, and getting_started/* use the older completion_config / agent_config API rather than the factory methods, so they don't depend on the 1.0.0 release.

async def async_main() was kept as-is in all three updated examples — each still awaits model.run() / agent.run() / graph.run() and response.evaluations, so the function must remain async.

End-to-end verification

Verified against the local SDK worktree (sync factory implementation) and re-verified that poetry install resolves launchdarkly-server-sdk-ai 1.0.0 from PyPI in all three examples after pinning.

• features/create_model (poetry run model with LAUNCHDARKLY_COMPLETION_KEY=sample-ai-config) — full chat + judge evaluation flow ran successfully.
• features/create_agent (poetry run agent with LAUNCHDARKLY_AGENT_KEY=sample-ai-config) — factory call succeeded and agent.run() returned a response. The example then errored on summary.usage because the example file uses .usage, but the SDK renamed that attribute to .tokens in 0.20 ("Rename LDAIMetrics.usage and AIGraphMetrics.usage to .tokens"). This is a pre-existing example bug unrelated to this PR and should be addressed separately. (See the same .usage references in features/create_agent_graph.)
• features/create_agent_graph — not run end-to-end. The change is mechanical and identical in shape to the other two factories, and the same pre-existing .usage issue would prevent a clean run.

Test plan

• CI passes.
• Spot-check one of the affected examples against the published 1.0.0 release.