Speed up airlfow roles list command

[potiuk]

The command initialized whole flask_app of ours, but what we really needed was FAB security manager only.

Fixes: #28242

---

^ Add meaningful description above

Read the Pull Request Guidelines for more information.
In case of fundamental code changes, an Airflow Improvement Proposal (AIP) is needed.
In case of a new dependency, check compliance with the ASF 3rd Party License Policy.
In case of backwards incompatible changes please leave a note in a newsfragment file, named {pr_number}.significant.rst or {issue_number}.significant.rst, in newsfragments.

[potiuk]

Maybe not the "FINAL" fix but on my slow local DB:

 time airflow roles list

Before:

name
======
Admin
Op
Public
User
Viewer

airflow roles list  5.49s user 1.79s system 136% cpu 5.339 total

After:

name
======
Admin
Op
Public
User
Viewer

airflow roles list  2.43s user 1.75s system 198% cpu 2.103 total

@malthe - can you check it on your system please ?

[malthe]

This cut our runtime from 49s to 9s.

[malthe]

Could this improvement be applied to roles_create as well?

[potiuk]

Could this improvement be applied to roles_create as well?

Very much so. I can do another PR

[potiuk]

There are quite a few other commands like that in user/role management that we could improve this way.

Hey @ashb @kaxil @jhtimmins and others - do you think there is any drawback to this solution ?

It does not initialize quite a lot of Airlfow Webserver pieces and we get 3x - 5x speedup for those commands.

Potentially I can imagine some subtle changes that somoene might want to add in their plugins might stop working with the CLI commands if we limit initialization to only Flask App Builder. However, if you have your own Airflow Security Manager - it should continue to work.

But I'd argue the CLIs of our was never meant to be extended by plugins, so I see that as "feature" not "bug/breaking change".

[potiuk]

BTW. I think we can speed it up even more by reaping apart some of the FAB code doing the intialization - but this is the "Pareto" rule in full swing - 80% of gain in 20% of the effort.

[potiuk]

@malthe - i sped it up a bit more - I suspected I do not even need "import_all_models" - so now it should be even faster.

[ashb]

LGTM.

[potiuk]

I will apply it elsewhere then :)