[meta] refactor scripts to stop importing scriptworker, start importing scriptworker_client

まず、scriptworker を import しているスクリプトを見つけ、コマンドラインのエントリポイントを scriptworker_client と比較します。Testing iscript wiki ページと、スタンドアロン実行のための create_test_workdir に関するガイダンスを読みます。完了の条件は、コマンドライン API を維持したまま、スクリプトが scriptworker のコンテキストオブジェクトに依存しなくなることです。

Part of the rationale of having scriptworker call a script from the commandline, rather than import a python module and call a function, was to make the script independent of scriptworker. In theory this would:

1. let us upgrade scriptworker without affecting scripts, as long as that commandline api stayed consistent,
2. allow us to test scripts standalone, rather than requiring the worker,
3. allow for a mixture of languages and python versions

(3) is currently no longer an issue. We've standardized on python 3. However, we still have the option of supporting other languages, especially if some other team wants to start using scriptworkers.

(2) has been theoretical, and still has some issues with user friendliness and discoverability. The testing iscript wiki page shows it's possible to test a script without scriptworker, especially when using the create_test_workdir script from scriptworker to populate the work_dir with a known task's task.json and CoT upstreamArtifacts.

The main thing we're solving here is (1). Because many scripts import scriptworker, upgrading scriptworker may still break those scripts, even if the commandline api stays consistent. Let's import scriptworker_client instead.

Doing so isn't just a simple drop-in replacement, however. scriptworker_client removes the usage of context objects, so this may be a sizeable patch set per script. But once we do this, we'll once more be able to upgrade scriptworker without fear of breaking the script, and we will be able to make all script-related changes in this repo rather than across two repos. (And the vocal opponents to the context objects will be appeased. :)