release-23.1: randgen: disable generation of REGNAMESPACE type expressions

[blathers-crl]

Backport 1/1 commits from #99168 on behalf of @msirek.

/cc @cockroachdb/release

---

Casting an OID to REGNAMESPACE runs this SQL using the internal executor:

SELECT pg_namespace.oid, nspname FROM pg_catalog.pg_namespace WHERE oid = $1

When the GenerateConstrainedScans rule is disabled, this returns no
rows and so doesn't apply the cast. The same SQL, not run via an
internal executor, but also with GenerateConstrainedScans disabled,
behaves correctly. Disabling GenerateConstrainedScans all the time
prevents the cluster from starting.

The reason a full scan of pg_namespace doesn't find the OID is because it
only checks for schemas in the current database:

cockroach/pkg/sql/virtual_schema.go

Line 602 in 7b341ff

dbDesc, err = p.byNameGetterBuilder().Get().Database(ctx, dbName)

cockroach/pkg/sql/pg_catalog.go

Lines 2099 to 2106 in d10c3dd

	var pgCatalogNamespaceTable = virtualSchemaTable{
	comment: `available namespaces
	https://www.postgresql.org/docs/9.5/catalog-pg-namespace.html`,
	schema: vtable.PGCatalogNamespace,
	populate: func(ctx context.Context, p *planner, dbContext catalog.DatabaseDescriptor, addRow func(...tree.Datum) error) error {
	return forEachDatabaseDesc(ctx, p, dbContext, true, /* requiresPrivileges */
	func(db catalog.DatabaseDescriptor) error {
	return forEachSchema(ctx, p, db, true /* requiresPrivileges */, func(sc catalog.SchemaDescriptor) error {

cockroach/pkg/sql/information_schema.go

Lines 2295 to 2296 in a7e9c4a

	// forEachDatabaseDesc calls a function for the given DatabaseDescriptor, or if
	// it is nil, retrieves all database descriptors and iterates through them in

Whereas with a constrained scan, all descriptors are searched:

cockroach/pkg/sql/catalog/descs/descriptor.go

Lines 168 to 175 in af6a72a

	for _, fn := range []lookupFunc{
	q.lookupVirtual,
	q.lookupTemporary,
	q.lookupSynthetic,
	q.lookupUncommitted,
	q.lookupCached,
	q.lookupLeased,
	} {

The correct result is from the constrained scan, because we currently
allow cross-database references. Once #55791 is fixed, both results
should match.

The fix alternatives are:

1. disallow disabling of the GenerateConstrainedScans rule for internal SQL
2. turn off generation of REGNAMESPACE expressions in randgen to avoid hitting this problem in tests.

The 2nd fix alternative is implemented to avoid losing any of the
rule-disabling test coverage provided by the
testing_optimizer_disable_rule_probability setting.

Fixes #98322

---

Release justification: Test-only temporary fix to avoid test flakes until #55791 is fixed.

[blathers-crl]

Thanks for opening a backport.

Please check the backport criteria before merging:

• Patches should only be created for serious issues or test-only changes.
• Patches should not break backwards-compatibility.
• Patches should change as little code as possible.
• Patches should not change on-disk formats or node communication protocols.
• Patches should not add new functionality.
• Patches must not add, edit, or otherwise modify cluster versions; or add version gates.

If some of the basic criteria cannot be satisfied, ensure that the exceptional criteria are satisfied within.

• There is a high priority need for the functionality that cannot wait until the next release and is difficult to address in another way.
• The new functionality is additive-only and only runs for clusters which have specifically “opted in” to it (e.g. by a cluster setting).
• New code is protected by a conditional check that is trivial to verify and ensures that it only runs for opt-in clusters.
• The PM and TL on the team that owns the changed code have signed off that the change obeys the above rules.

Add a brief release justification to the body of your PR to justify this backport.

Some other things to consider:

• What did we do to ensure that a user that doesn’t know & care about this backport, has no idea that it happened?
• Will this work in a cluster of mixed patch versions? Did we test that?
• If a user upgrades a patch version, uses this feature, and then downgrades, what happens?

[cockroach-teamcity]

This change is 

[mgartner]

LGTM.

Don't forget to post to #release-backports and add a release justification.

[msirek]

backport approval request here

[msirek]

TFTR!