fix(schema): use anyOf where BNF allows mixed inline and named groups

[aorzelskiGH]

Summary

Replaces oneOf with anyOf in exactly those places where the BNF
uses concatenation (both variants may appear together) rather than
alternation (exactly one). Keeps oneOf where the BNF is genuinely
XOR.

Problem

The BNF makes a deliberate distinction between "mix allowed" and
"exactly one":

<AttributeGroup> ::=
    ( <SingleAttribute> <ws> )*
    ( <UseAttributeGroup> <ws> )*     -- CAT (mix allowed)

<AccessPermissionRule> ::=
    ( <ACL> | <UseACL> )                -- XOR
    "OBJECTS:" <ws>
    ( <SingleObject> <ws> )*
    ( <UseObjectGroup> <ws> )*          -- CAT (inline)
    ( "FORMULA:" ... | <UseFormula> )   -- XOR

But the JSON Schema used oneOf for the CAT pairs too, making the
schema strictly more restrictive than the grammar. A rule that lists
both inline ATTRIBUTES and a USEATTRIBUTES reference was legal
per BNF and illegal per schema.

Solution

Change oneOf to anyOf in exactly the CAT cases:

• ACL.{ATTRIBUTES, USEATTRIBUTES}
• AccessPermissionRule.{OBJECTS, USEOBJECTS}
• aas-specs-security only: DEFATTRIBUTES item
  {attributes, USEATTRIBUTES}

Keep oneOf for the XOR cases:

• AccessPermissionRule.{ACL, USEACL}
• AccessPermissionRule.{FORMULA, USEFORMULA}
• DEFOBJECTS item {objects, USEOBJECTS} (grammar
  <ObjectGroup> is XOR: ( A )* | ( B )*)
• SecurityQueryFilter.{CONDITION, USEFORMULA}

Impact

• Affected specs: IDTA-01002 (API) and IDTA-01004 (Security).
• Strict relaxation: every document previously valid remains valid;
  additionally, documents that combine inline and named attribute/
  object groups now validate.
• No grammar changes.

Review notes

• Please confirm the ObjectGroup grammar is indeed intended as XOR
  (production line: ( <SingleObject> )* | ( <UseObjectGroup> )*)
  while the inline OBJECTS: block inside <AccessPermissionRule>
  is intended as CAT. If both are meant to be CAT, DEFOBJECTS
  should also be flipped to anyOf — trivial follow-up.
• The single-string USEATTRIBUTES in ACL vs the array-of-string
  in DEFATTRIBUTES is a separate inconsistency and is not
  addressed here.

Related

Review Finding T-13: XOR vs mix semantics between BNF and JSON
Schema.