Ensure Normalize func preserves milliseconds precision (#423)

**Description:**
Fixes an issue in the OS STAC `/search` endpoint where datetime filters
overwrite milliseconds. The queries return products outside the
specified range because the filter was rounding datetimes to full
seconds. This change ensures that milliseconds are preserved, so
datetime filtering returns correct results.

**PR Checklist:**

- [x] Code is formatted and linted (run `pre-commit run --all-files`)
- [x] Tests pass (run `make test`)
- [ ] Documentation has been updated to reflect changes, if applicable
- [x] Changes are added to the changelog

---------

Co-authored-by: Yuri Zmytrakov <[email protected]>

Author: YuriZmytrakov

CHANGELOG.md

@@ -22,6 +22,7 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 
 - Added `id` field as secondary sort to sort config to ensure unique pagination tokens. [#421](https://github.com/stac-utils/stac-fastapi-elasticsearch-opensearch/pull/421)
 - Added default environment variable `STAC_ITEM_LIMIT` to SFEOS for result limiting of returned items and STAC collections [#419](https://github.com/stac-utils/stac-fastapi-elasticsearch-opensearch/pull/419)
+- Updated the `format_datetime_range` function to support milliseconds. [#423](https://github.com/stac-utils/stac-fastapi-elasticsearch-opensearch/pull/423)
 
 ### Changed
 

stac_fastapi/core/stac_fastapi/core/datetime_utils.py

@@ -17,17 +17,20 @@ def format_datetime_range(date_str: str) -> str:
     """
 
     def normalize(dt):
+        """Normalize datetime string and preserve millisecond precision."""
         dt = dt.strip()
         if not dt or dt == "..":
             return ".."
         dt_obj = rfc3339_str_to_datetime(dt)
         dt_utc = dt_obj.astimezone(timezone.utc)
-        return dt_utc.strftime("%Y-%m-%dT%H:%M:%SZ")
+        return dt_utc.isoformat(timespec="milliseconds").replace("+00:00", "Z")
 
     if not isinstance(date_str, str):
         return "../.."
+
     if "/" not in date_str:
         return f"{normalize(date_str)}/{normalize(date_str)}"
+
     try:
         start, end = date_str.split("/", 1)
     except Exception: