[BUG] matrix::select_k: inconsistent handling of infinities

[achirkin]

matrix::select_k uses a heuristic to select one among several implementations depending on the input arguments.
The problem is that some of the implementation have a special handling of max_bound/min_bound (inifinity) values; they may replace the corresponding indices (payload) with special values, such as zero or numeric_limit<T>::max().
This problem manifests itself when the number of valid (non-infinity) values is less than k. Example: IVF-PQ may return many infinities if the low-precision internal types are used; if the refinement used afterwards, it can return the zeroth vector from the database multiple times.
This can be tested in the newly added tests:

raft/cpp/test/matrix/select_k.cuh

Lines 240 to 243 in ea160fa

	// Some algorithms return invalid indices in special cases.
	// This can be considered as TODO for us to fix.
	if (static_cast<size_t>(ix_i) >= in_ids.size()) return forgive_i;
	if (static_cast<size_t>(ix_j) >= in_ids.size()) return forgive_j;

This listing summarizes what indices (payloads) do different implementation return:

raft/cpp/test/matrix/select_k.cuh

Lines 258 to 274 in ea160fa

	auto forgive_algo(const std::optional<select::Algo>& algo, IdxT ix) const -> bool
	{
	if (!algo.has_value()) { return false; }
	switch (algo.value()) {
	// not sure which algo this is.
	case select::Algo::kPublicApi: return true;
	// warp-sort-based algos currently return zero index for inf distances.
	case select::Algo::kWarpAuto:
	case select::Algo::kWarpImmediate:
	case select::Algo::kWarpFiltered:
	case select::Algo::kWarpDistributed:
	case select::Algo::kWarpDistributedShm: return ix == 0;
	// FAISS version returns a special invalid value:
	case select::Algo::kFaissBlockSelect: return ix == std::numeric_limits<IdxT>::max();
	// Do not forgive by default
	default: return false;
	}