feat(base-query): Indexable(each = true) with contains/doesNotContain

[deer]

• @Indexable gains an each() attribute (default false). When each = true, the function is expected to return a Stream, Collection, or Iterable, and each element is indexed individually as a separate reverse-map key. When each = false (the default), the return value is treated as a single scalar key, exactly as before. Intent is declared at the field rather than inferred at runtime from the return type, so a function returning List<Tag> is never silently treated as multi-valued unless the author explicitly opts in.
• AbstractHeapBasedIndex routes scalar and each functions through entirely separate code paths (indexNonUnique/indexEach, reindexNonUnique/reindexEach) backed by two new memoizers. The previous putNonUniqueValue with its runtime instanceof dispatch is removed. If an each = true function returns a non-container value, it throws at index time with a clear message.
• Condition gains contains(Object element) for O(1) indexed membership lookup (falling back to a linear scan for non-indexed functions) and doesNotContain(Object element) which always scans. A matches-based default was considered for doesNotContain and rejected: Matches.findAll() iterates over individual reverse-map keys rather than the original container, which would return wrong results for indexed each-functions.
• shouldDistinguishScalarIndexingFromEachIndexing puts both modes side by side on the same List<Color> field -- COLORS_SCALAR with isEqualTo(list) and COLORS_EACH with contains(element) — making the difference concrete and executable. Additional contract tests cover Collection and Iterable element types, reindex after mutation, the non-indexed fallback path, and doesNotContain exclusion semantics.