Improve ergonomics of data migration test

[smklein]

• Splits up each data migration into a separate file
• Uses migration name instead of migration semver version, to reduce conflicts

This PR should make it easier to concurrently merge data migrations, but it also should make it easier to "remove old data migrations", identified by #8271.

[smklein]

FYI! Of all the files in data_migrations.rs to read, this is the most important one to review.

The others are "Just moving" test functions out of schema.rs

[hawkw]

This is awesome. I had a couple small suggestions, but I think this will make writing these tests much less annoying!

[hawkw]

nit, take it or leave it: since register is evaluated inside this function exclusively, i think the map and versions identifiers don't actually need to be passed in?

Suggested change

	macro_rules! register {
	($map:ident, $versions:ident, $mod:ident) => {
	let name = stringify!($mod).replace('_', "-");
	let version = $versions
	.get(name.as_str())
	.unwrap_or_else(|| {
	panic!("migration {name:?} not found in KNOWN_VERSIONS")
	})
	.clone();
	$map.insert(version, $mod::checks());
	};
	}
	pub(crate) fn get_migration_checks() -> BTreeMap<Version, DataMigrationFns> {
	let versions: HashMap<&str, Version> = KNOWN_VERSIONS
	.iter()
	.map(|v| (v.relative_path(), v.semver().clone()))
	.collect();
	let mut map = BTreeMap::new();
	register!(map, versions, zone_image_resolver_inventory);
	register!(map, versions, vpc_firewall_icmp);
	register!(map, versions, boot_partitions_inventory);
	register!(map, versions, fix_leaked_bp_oximeter_read_policy_rows);
	register!(map, versions, route_config_rib_priority);
	register!(map, versions, inv_clear_mupdate_override);
	register!(map, versions, populate_db_metadata_nexus);
	register!(map, versions, positive_quotas);
	register!(map, versions, disk_types);
	register!(map, versions, one_big_ereport_table);
	register!(map, versions, blueprint_sled_config_subnet);
	register!(map, versions, blueprint_sled_last_used_ip);
	register!(map, versions, audit_log_credential_id);
	register!(map, versions, fix_session_token_column_order);
	register!(map, versions, bgp_unnumbered_peers);
	register!(map, versions, bgp_config_max_paths_not_null);
	register!(map, versions, ereport_everyone_gets_a_slot);
	register!(map, versions, rename_default_igw_ip_pool);
	macro_rules! register {
	($mod:ident) => {
	let name = stringify!($mod).replace('_', "-");
	let version = versions
	.get(name.as_str())
	.unwrap_or_else(|| {
	panic!("migration {name:?} not found in KNOWN_VERSIONS")
	})
	.clone();
	map.insert(version, $mod::checks());
	};
	}
	pub(crate) fn get_migration_checks() -> BTreeMap<Version, DataMigrationFns> {
	let versions: HashMap<&str, Version> = KNOWN_VERSIONS
	.iter()
	.map(|v| (v.relative_path(), v.semver().clone()))
	.collect();
	let mut map = BTreeMap::new();
	register!(zone_image_resolver_inventory);
	register!(vpc_firewall_icmp);
	register!(boot_partitions_inventory);
	register!(fix_leaked_bp_oximeter_read_policy_rows);
	register!(route_config_rib_priority);
	register!(inv_clear_mupdate_override);
	register!(populate_db_metadata_nexus);
	register!(positive_quotas);
	register!(disk_types);
	register!(one_big_ereport_table);
	register!(blueprint_sled_config_subnet);
	register!(blueprint_sled_last_used_ip);
	register!(audit_log_credential_id);
	register!(fix_session_token_column_order);
	register!(bgp_unnumbered_peers);
	register!(bgp_config_max_paths_not_null);
	register!(ereport_everyone_gets_a_slot);
	register!(rename_default_igw_ip_pool);

which might be slightly nicer for the user?

[smklein]

Sure, updated. Gotta move the macro inside the function for this to work, but it does (done in faf7e1b)

[hawkw]

I think it might be nice if this explained a bit more about what "advancing the schema baseline" means. When would I need to do this?

[smklein]

Explained things a bit, and linked to docs in faf7e1b