move favor state arg to click

[ChenyuLInx]

resolves #6585

Tests related to --favor-state passes. There are still two fail but not related to --favor-state. Planning on resolving those at the very end.

tests/functional/defer_state/test_defer_state.py ....FF..                [100%]

[github-actions]

Thank you for your pull request! We could not find a changelog entry for this change. For details on how to document a change, see the contributing guide.

[jtcohen6]

@ChenyuLInx I figured it out! The tests fail for me locally, and then pass when I switch -m to -s in the arguments to run_dbt (83850f5)! It seems like this might be a genuine bug where we're not passing through --models/-m appropriately. That's important for backward compatibility.

Desirable behavior:

• For almost all commands, --models should be a pure alias for --select (and the two should be mutually exclusive)
• EXCEPTION: dbt list historically accepted both --select and --models, with slightly different behavior, which we've preserved for back compat

---

As a separate matter, docs generate --defer is indeed a bit of an odd duck:

• docs generate --no-compileis effectively incompatible with--defer, because docs generaterelies on thecompile` step to perform the manifest state comparison/merge
• Selection criteria for docs generate only currently impacts which nodes are compiled, not which nodes are catalogued. (We might want to change that: [CT-1303] Respect node selection in catalog queries run by docs generate #6014)
• Personally, I'd imagine wanting to catalog whichever version of the node actually exists—my dev version if it exists, the prod version if it doesn't—ignoring selection logic entirely (24872f1)

That's obviously out of scope for this PR

[aranke]

LGTM, one naming nit.

[aranke]

Nit: this should be defer_favor_state or simply favor_state.

[stu-k]

Agreed.