Add experimental cache_selected_only config

[jtcohen6]

resolves #4688
rebased version of #4860
more tracking + testing to come in #4961

Only other changes:

• --cache-selected-only (kebab case) instead of --cached_selected_only (snake case)
• Add a breaking change note for the internal adapter method (Python code) with a change in its signature. Among our plugins, only dbt-postgres reimplements this method; all the others inherit the base version.

I'd love to include this as "experimental" functionality in v1.1 if possible. We would document it as such: dbt-labs/docs.getdbt.com#1327

Checklist

• I have signed the CLA
• I have run this code in development and it appears to resolve the stated issue
• This PR includes tests, or tests are not required/relevant for this PR
• I have added information about my change to be included in the CHANGELOG.

[gshank]

Looks good! I see there's only a test for the flag setting not the caching though.

[jtcohen6]

@gshank You're right :) This is definitely going to be "experimental" / "use at your own risk," but worth at least one test. I did a bad thing, and just extended an old test for caching, rather than converting.

[gshank]

Changing an existing test is perfectly legit :) They ought to behave the same...

[ChenyuLInx]

Late to the party but this looks great!

[rmargarint-nydig]

I realized this is a comment on an old PR, but I was experimenting today with cache_selected_only to speed up start up time on our test execution pipelines which run as part of Airflow orchestration. It seems that cache_selected_only only seems to affect dbt run but not dbt test --select "test_name"? From what I've traced, it seems get_model_schemas(...) will return an empty set when one of the selected uids is a test resource, which will result in the entire set of schemas being pulled into the cache. Is this expected?