Vars invalidation (#2265)

[beckjake]

resolves #2265

Description

On RPC calls, if the server detects that the final CLI vars calculated from --vars are different from the server's CLI --vars, rebuild the manifest.

This will make those calls slower, because they have to build a manifest (and the partial parsing cache will be invalidated!).

I also moved the vars calculation from set_config (which happens in the main dispatching process) into handle_request (which happens in the RPC process). That fixed an issue where future calls had the wrong vars once you made a call using --vars (ugly!)

I also split up the RPC tests, that file was getting crowded. That accounts for the vast majority of this change, I've noted the only new/changed test.

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 updated the CHANGELOG.md and added information about my change to the "dbt next" section.

[beckjake]

This is the only new test in this PR.

[drewbanin]

This looks really good to me! I'm not so worried about a performance impact here - this should only affect cli_args method calls which don't need to be incredibly snappy (as compared to something like compile_sql or run_sql, for instance).

🚢 🚢 🚢