Normalize initial database state for new installations

[anonymoususer72041]

Summary

This PR normalizes the initial database state created from db/cats_schema.sql by removing demo/sample tag data, resetting instance-specific seeded values to deterministic defaults, and replacing testdomain.com placeholders with the IANA-reserved placeholder domain example.com.

In addition, it updates modules/install/Schema.php to migrate existing databases where site.last_viewed_day is NULL or 0000-00-00 by backfilling a sentinel date (1000-01-01) and then enforcing a non-null default at the schema level.

Motivation

Fresh installs should start from a clean, deterministic baseline without embedded demo content or leftover environment-specific values. This reduces confusion, makes the initial schema easier to reason about, and avoids noisy diffs when comparing or exporting databases.

Backfilling and normalizing site.last_viewed_day prevents NULL/zero-date edge cases and aligns this field with the sentinel-date approach used elsewhere in the schema while keeping existing instances consistent through an explicit migration step.

Replacing testdomain.com with the IANA-reserved placeholder domain example.com avoids potential confusion with real domains.

[anonymoususer72041]

@RussH any chances to merge this PR short term?

I’m currently working on a follow-up PR that removes test.sql to reduce drift and the ongoing effort of maintaining two separate databases. I want to avoid another stacked PR.

[anonymoususer72041]

@RussH CI/CD “Tests (PHP 7.2)” is now failing again, and it’s the exact drift issue I mentioned around test.sql.

Behat fails at the very first “Login as administrator” step (“Invalid username or password”), which causes the rest of the scenarios to cascade (missing fields/pages because the suite is stuck on the login screen). The current test/data/test.sql seeds the admin password in a way that no longer matches what the app expects (hashing), so the credentials in test/features/login.feature can’t authenticate.

This is another concrete example of why maintaining both cats_schema.sql and test.sql as separate sources keeps breaking CI whenever one side changes. As a short-term band-aid, I’ll push a minimal fix to re-align the seeded admin password in test.sql with the expected hashing scheme so CI is green again.

After that, I’d really like to start on the follow-up PR that removes test.sql (or replaces it with a generated/derived fixture) to prevent this from recurring.