Clean up fabric filtered read

[mlepage-google]

See PR #13881 and particularly this conversation:
#13881 (comment)

Essentially, if not using the generated cluster types for encode, they won't fabric filter if a MatchesFabricIndex method isn't provided. They also won't fail to compile or otherwise warn, so this is highly dangerous.

There is discussion in that conversation about various ways to address this. It should be addressed at the DM/IM level as appropriate, and then whatever changes need to happen to access control cluster (for whatever new mechanism is used for fabric filtered read) will have to happen.

[erjiaqing]

I guess this is resolved,

IsFabricScoped now requires "kIsFabricScoped" attribute always exists when encoding structs.

It says

 * Using IsFabricScoped<X>::value without a basic type or cluster object with kIsFabricScoped will cause a compile error.  This is
 * an intended behavior to make users explicitly decide whether their cluster object is fabric-scoped or not.

connectedhomeip/src/app/data-model/FabricScoped.h

Lines 36 to 54 in 87d1d3e

	template <typename X>
	class IsFabricScoped
	{
	private:
	template <typename X0, decltype(std::decay_t<X0>::kIsFabricScoped) = true>
	static constexpr bool IsFabricScopedClusterObject()
	{
	return std::decay_t<X0>::kIsFabricScoped;
	}
	template <typename X0, std::enable_if_t<IsBasicType<std::decay_t<X0>>::value, bool> = true>
	static constexpr bool IsFabricScopedClusterObject()
	{
	return false;
	}
	public:
	static constexpr bool value = IsFabricScopedClusterObject<X>();
	};