Revise `agency_source_link` and remove `agencies.data_sources`

[maxachis]

Context

• We previously developed an agency_source_link table. However, this table was created in production and hence we don't have a clear record of how it was populated, and whether that population is kosher.
• We also still have the agencies.data_sources column, which means we currently have two sources of truth for the link between agencies and data sources.

Requirements

• [x] Rebuild the agency_source_link table such that its results are in dev_scripts.sql.
• [x] Remove the agencies.data_sources column.

Follow-up:

• [ ] #460
• [ ] #461
• [ ] #458
• [ ] #459

Tests

• All existing tests should still function as expected.

Docs

• Add comments to agency_source_link describing the table and what each column does.

Open questions

• Any uncertainties? Anything we're flexible about?

[josh-chamberlain]

agencies.data_sources is one of those magic multi-reference fields in Airtable, and it strikes me as possible that the agency_source_link table was built that way. Is the idea to build the link table instead by checking each source for its agencies? If so...I approve

[maxachis]

@josh-chamberlain I'm also planning to remove the other column that reference these, such as data_sources_last_updated and data_sources_count (which are stale, anyway).

I'll also note that this breaks /homepage-search-cache, so I'm commenting out that endpoint and its tests and creating a new issue for fixing that (if we want to):