Stabilize CI

[mgravell]

1. ensure we have healthchecks so up --wait works correctly and we don't (as a race condition) start against an incomplete cluster
2. don't assert FT.INFO internal metrics - just log them; they drift too much between module versions and are dependent on GC timing
3. add indexing pause in QueryEmExample

---

Note

Low Risk
Test and CI infrastructure only; no production library or runtime behavior changes.

Overview
Reduces flaky CI by fixing when tests start against Redis and what they assert after search indexing.

Docker Compose adds healthchecks on standalone (PING) and cluster (cluster_state:ok) so docker compose up --wait does not proceed until Redis accepts commands and the cluster has all slots assigned—avoiding races that produced spurious -MOVED from stale slot maps.

Search tests replace brittle FT.INFO internal statistic assertions (version- and GC-sensitive) with logging on standalone, while keeping correctness checks (NumDocs, HashIndexingFailures, PercentIndexed, etc.).

Doc example QueryEmExample inserts 2s sleeps (stripped from published docs) after JSON writes so async RediSearch indexing can catch up before FT.SEARCH assertions under load.

^{Reviewed by Cursor Bugbot for commit b1ec7ed. Bugbot is set up for automated code reviews on this repo. Configure here.}