bug 1889156: script to load processed crashes into Elasticsearch

[biancadanforth]

Because:

• We need to migrate Elasticsearch from AWS to GCP as part of the Socorro GCP migration.
• Due to our version of Elasticsearch being very old, the best (only?) possible migration path is to load data into the GCP Elasticsearch cluster from the original source, which is S3 in our case.

This PR:

• Adds a script that loads processed crashes from either S3 or GCS (depending on settings.CRASH_SOURCE) into Elasticsearch by either date (YYYY-MM-DD) or crash ID.

How to test this locally

I have painstakingly added test cases for all the major scenarios, so running the tests in socorro/tests/test_load_processed_crashes_into_es.py (or replicating the setup before running the script with similar args) should be sufficient.

The test cases conditionally skip based on settings.CLOUD_PROVIDER, so to test the script with GCS as the crash storage source locally, you can edit your my.env so that CLOUD_PROVIDER=GCP and rebuild the test container (which in my case is the devcontainer). Just don't forget to revert that change when you're done. See #6609 (comment) -- this PR now runs both test cases against both cloud providers thanks to some clever pytest config.

I can add some manual testing instructions if things aren't clear from the test. I might just add those here tomorrow.

[biancadanforth]

I couldn't think of any special types of exceptions that we'd want to handle differently -- are there any?

Also, please let me know if we'd like to handle exceptions in the general case differently -- right now we just log to stdout that we couldn't load it and move on.

[biancadanforth]

Okay I've updated the PR for all your feedback except for #6609 (comment) . I'll submit a separate commit for that tomorrow, so it's easier to revert if willkg weighs in and says it's a bad idea.

[relud]