♻️ Replace aiopg in catalog service

[sanderegg]

What do these changes do?

This PR is following #2864 that updated SQLalchemy to the latest version and left the catalog service with some functional but suboptimal code such that aiopg could work with the newer SQLalchemy syntax.

This PR replaces aiopg/psycopg2 with asyncpg and allows a full integration with SQLalchemy ONLY in the catalog for now.
This PR also introduces the use of the library aiocache and is currently used to cache the list of services returned by the director-v0 for a small performance boost in listing the services according to service access rights. This library can also be used with redis to have a distributed cache.
Currently we list all the services with all the details, etc. The most time consuming part is creating the ServiceOut objects at the end. There I don't see how to fix that for now, unless we reduce the number of objects?

Using the new async engine in SQLalchemy implies:

• replace any occurence of acquire() with connect()
• if using a construction with async for row... then one needs to call conn.stream( instead of conn.execute(
• SQL calls with a string is not allowed, one needs to pass a sa.DDL type with the string inside

Additional:

• added a fixture to create a SQLalchemy async engine in postgres_service.py
• catalog now caches the list of services in the director-v0 for 5 minutes using a TTL cache

Bonus:

• added pytest-benchmark for benchmarking tests

reference for SQLAlchemy in async mode: https://docs.sqlalchemy.org/en/14/orm/extensions/asyncio.html#synopsis-core

Related issue/s

How to test

Checklist

[codecov]

Codecov Report

Merging #2869 (66fcd7e) into master (6c63bef) will decrease coverage by 6.3%.
The diff coverage is 61.1%.

@@           Coverage Diff            @@
##           master   #2869     +/-   ##
========================================
- Coverage    79.1%   72.8%   -6.4%     
========================================
  Files         673     674      +1     
  Lines       27602   27620     +18     
  Branches     3218    3220      +2     
========================================
- Hits        21854   20117   -1737     
- Misses       5001    6856   +1855     
+ Partials      747     647    -100     

Flag	Coverage Δ
integrationtests	?
unittests	72.7% <61.1%> (-2.0%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...c/simcore_postgres_database/utils_aiosqlalchemy.py	0.0% <0.0%> (ø)
...catalog/src/simcore_service_catalog/core/events.py	90.0% <ø> (ø)
...imcore_service_catalog/db/repositories/projects.py	41.6% <0.0%> (ø)
...rc/simcore_service_catalog/db/repositories/dags.py	44.4% <11.1%> (-4.3%)	⬇️
.../settings-library/src/settings_library/postgres.py	86.8% <50.0%> (-4.4%)	⬇️
...g/src/simcore_service_catalog/services/director.py	76.3% <50.0%> (ø)
.../simcore_service_catalog/db/repositories/groups.py	72.9% <70.0%> (+16.2%)	⬆️
...imcore_service_catalog/db/repositories/services.py	77.7% <76.4%> (+5.5%)	⬆️
...s/catalog/src/simcore_service_catalog/db/events.py	75.0% <87.5%> (-10.8%)	⬇️
...mcore_service_catalog/api/dependencies/database.py	100.0% <100.0%> (+12.5%)	⬆️
... and 113 more

[GitHK]

👍 very nice

[GitHK]

I guess this will stay in there until we fully remove aiopg. Maybe put a note.

[sanderegg]

this is the only location where the async engine is created (pytest-simcore package). but no worries, the PostgresSettings also have a special attribute that does it for you.

[pcrespov]

Very nice work. Just a few comments!

[pcrespov]

two things, we already introduce this "extra"

sqlalchemy[postgresql_psycopg2binary]

should we always extra with both?

sqlalchemy[asyncio,postgresql_psycopg2binary]

[sanderegg]

no I think once we can fully migrate, the psycopg2binary becomes obsolete. maybe just for testing where we do have a sync engine as well