[Artifacts] Improve list artifacts querying

[TomerShor]

Previously, when listing artifacts we did several intensive operations against the DB:

1. When asking a specific tag, we ran and SQL query on the tags, than filtered the artifacts on their UIDs from the returned tags.
2. For adding tags to the artifact object, we:
   1. Queried the db for all artifacts with the given filters
   2. Iterated over the returned artifacts
   3. For each artifact, query the DB for its tags and add them to the struct.

All of the above resulted in a lot of overhead and too many queries.

Now we:

1. Query all artifacts, joined* on the artifact tags, filtered by tag name. (This will already return the permutations)
2. For each artifact add the tag to its metadata.
3. Moved the producer_uri filtering to the same for loop as (2) to not iterate over all artifacts again.

*outer join if no specific tag was requested so we return artifacts that are tag-less, and a simpler join when querying on a specific tag as in that case tag-less functions are irrelevant.

https://iguazio.atlassian.net/browse/ML-6642

[quaark]

Looks good! Couple of things

[quaark]

One more minor thing

[quaark]

👑