Validate inputs of TypeResolver::resolveExpression(): only allow %parent, %type and %key

1. +++ b/core/lib/Drupal/Core/Config/Schema/TypeResolver.php
   @@ -84,8 +84,17 @@ private static function replaceVariable(string $value, $data) {
   +    $previous_name = null;
   

   s/null/NULL/

2. +++ b/core/lib/Drupal/Core/Config/Schema/TypeResolver.php
   @@ -84,8 +84,17 @@ private static function replaceVariable(string $value, $data) {
   +        throw new \LogicException('The only valid usages of a config schema with a % in it are %parent, %key, and %type in ' . $value);
   

   \Drupal\Core\Config\TypedConfigManager::replaceVariable() calls these variable values, we should probably do the same here.

3. +++ b/core/lib/Drupal/Core/Config/Schema/TypeResolver.php
   @@ -84,8 +84,17 @@ private static function replaceVariable(string $value, $data) {
   +        throw new \LogicException('%type can only used when immediatly preceeded by %parent in ' . $value);
   

   Typos 😅

This would complement #3406487: Consistently use "dynamic type name" and "expression" instead of "variable value" in TypedConfigManager's terminology very well.

#3406487: Consistently use "dynamic type name" and "expression" instead of "variable value" in TypedConfigManager's terminology landed!

@borisson_: Could we add a secondary MR here that does not modify TypeResolver (which does not exist yet, which is why this issue is postponed), but instead modifies TypedConfigManager? We'd have to rebase the EntityBundleExists MR that this is blocked on, but we probably need to rebase it anyway at this point.

#3382581: Add new `EntityBundleExists` constraint landed so this is now unblocked!

Transformed #5 into a MR. Now reviewing 🤓

FWIW, this is becoming more important, because other issues are starting to adopt TypeResolver too — for example #3324140: Convert field_storage_config and field_config's form validation logic to validation constraints.

Looking close, two small pieces of feedback 😊

@borisson_ You'll want to modify the test logic to not just mock a parent level, but a grandparent level. That way you'd be able to at least test %parent.%parent.entity_type_id, and then there's >1 test case again :)

Was going to RTBC, then spotted one last small problem. 🫣

Updated the issue summary for you :)