Mlrun

8d1ea445 [Datastore] Support writing partitioned parquet data (898)
e0c11296 [Feature store] Fix trying to read data when it is None (945)
a38b6af4 [Runtimes] Support code archives, enhance run cli and add CI/CD support and docs (888)
17fd21e6 [Runtimes] Fix auto_mount precedence to PVC->V3IO (944)
1b6831ed [Doc] Add feature store training and serving docs (927)

ef051f8b [Requirements] Bump nuclio-jupyter minimum to 0.8.13 (943)
b90fd207 [Sparkjob] Remove synchronous monitoring in job submission (941)
5e98291b [API] Fixes for exact match in Artifact lists (942)
89c94caa [Feature store] service_online_feature.get() will return columns with None value (940)
06bc6425 [Feature store] Change default targets path (934)
b4a981f2 [Builder] Use `python -m pip` instead of `pip` (best practice) (939)
793b9ce9 [SDK] Make auto_mount importable from root (938)
b9c73ff3 [Jupyter image] Fix installations to be done by notebook user (937)
84b22fc5 ci: prevent race condtion in unstable releases
42a084f6 [API] Enable to enrich function with node selection attributes on job submission endpoint (935)
33d6192b [Jupyter image] Fix API exploding trying to connect to itself + other fixes (848)
a8a438a0 [Feature store] Support deployment of an empty graph ingestion service (932)
5901bf04 [Doc] Add feature set doc + some renamings(931)
b473c0b7 [Feature store] Set default cache flush interval to 5 minutes (933)
11c94940 [Feature store] Adding option to specify uid/tag in delete methods (920)
cb1633ac [Doc] Adding documentation for storey transformation functions (913)
15580046 [Projects] Support being a follower (839)

a3907771 [Images] Refactor to have hierarchical structure (873)

Main features & enhancements
* **Secret Store**: Support azure key store, theSaarco
* **Model Monitoring**: Add endpoint registration to v2 serving + config changes, Michaelliv
* **Artifacts**: Support filtering by iteration in list artifacts, theSaarco
* **Feature store**: Set default flush interval in storey's cache, dinal
* **Feature store**: Handling multiple targets of same type/name/path, urihoenig
* **Feature store**: Support multiple ingestion engines (storey, pandas, spark), yaronha
* **Feature store**: Support filtering by date in ingest, katyakats
* **Feature store**: Support EmitPolicy in aggregations, katyakats
* **Doc**: Add how-to notebook for model monitoring deployment, Michaelliv
* **Doc**: Add feature set transformations documentation, theSaarco
* **API**: Add ingest endpoint to feature store, theSaarco
* **DB**: Change default identifier query to be exact match and support ~ prefix for doing ilike match, theSaarco
* **Sparkjob**: Block default gpus requests interface, urihoenig
* **Dask**: Enable to set resources for scheduler/worker separately, hedingber
* **UI**: [Release notes](https://github.com/mlrun/ui/releases/tag/v0.6.3)

More info can be found in the RCs release notes:

847ab9d9 [Images] Use base python images from quay.io/mlrun and pin specifically to 3.6.12 and 3.7.9 (926)
abc5431c [Feature store] Fix wrong handling of aggregated features in NoSql target (924)
e05ccb5b [Doc] Update Model monitoring docs (895)

7a0eab01 [DB] Change default identifier query to be exact match and support ~ prefix for doing ilike match (923)