Dry Run Protocol

[achirkin]

The dry run protocol defines a mechanism to simulate the execution of algorithms to get a precise estimate of the memory requirements for a real execution with the same parameters.

#include <raft/util/dry_run_memory_resource.hpp>

raft::resources res;
// auto my_function(const raft::resources& res, my_args...);
auto stats = raft::util::dry_run_execute(res, my_function, my_args...);
// stats.device_global_peak  – peak device memory (bytes)

This PR:

• Introduces new infrastructure:
  1. raft::util::dry_run_execute, tracking memory resource, resource::get_dry_run_flag) that lets callers estimate peak memory usage of any RAFT algorithm without executing GPU work.
  2. resource::pinned_memory_resource, resource::managed_memory_resource - so that all memory resources available in raft are bound to the associated raft::resources handle and can be temporarily replaced.
  3. breaking change unified host and pinned mdarray policies to be the host policy using different std::pmr resources. This change is hidden behind a few layers of types in the mdarray template arguments, so none but most exotic use cases should be affected.
• Makes all public functions across all raft namespaces dry-run compliant: allocations are always visible to the tracker; CUDA work is skipped.
• Adds a small user guide (docs/source/dry_run_protocol.md)