[PR #865/de362160 backport][stable-3] postgresql_db: add session_role support for raw connections

[patchback]

This is a backport of PR #865 as merged into main (de36216).

SUMMARY

Fixes #864

Fixes postgresql_db "session_role" behavior when state is dump/restore for binary (.pgc) archives.
It was being ignored but now it behaves as intended by pg_dump/pg_restore

ISSUE TYPE

• Bugfix Pull Request

COMPONENT NAME

postgresql_db

ADDITIONAL INFORMATION

When pg_dump is used, the --role parameter is used to switch to a different user/role that has sufficient permissions to perform the dump.

When pg_restore is used, the --role parameter, in combination with --no-owner, allows you to restore the objects contained in the archive under new ownership (the one specified with --role).

Up to now, whenever pg_dump/pg_restore was used, the session_role parameter was being ignored and it would give surprising results (or errors).

Now the session_role below will be relayed as the --role parameter for pg_dump/pg_restore.

- name: pg_restore archive on dst_db with session_role
  postgresql_db:
    login_user: postgres
    name: dst_db
    state: restore
    target: /tmp/archive.pgc
    session_role: restoration_role
    target_opts: "--no-owner"

I chose to create a new test python module for this feature, as it seemed cleaner and less intrusive with regards to the current integration tests.

[Andersson007]

ignoring failures against the Remote targets we are removing in #904

[Andersson007]

@gspanos hi, this is a backport PR to version 3, I see integration test failures, would you like to take a look?
I won't include it in 3.14.3 I'm releasing right now, but we can do it for the next one if we tackle the failures

[gspanos]

@Andersson007 sure, I'll take a look

[Andersson007]

@Andersson007 sure, I'll take a look

@gspanos great thanks! Please ignore RH 8.8 and RH 9.3 failures: we've removed those checks as obsolete in another PR, FYI