Performance: potential micro-optimisation on tags and local/remote sessions

Explanation
I'm profiling the C++ SDK on KV and transactions operations currently and seeing a few potential micro-optimisations that could be made. Take or leave these thoughts as you will I'll file them separately.

The source for the test app is here - note that only run_kv_lookup_in_workload is called, and that this workload simulates one of the reads that transactions do under-the-hood. The profiler is MSVC 2022.

"2% of TT" means 2% of the total CPU time as reported by the profiler. It's a sampling profiler using CPU clock not wallclock.

The issue

2% of TT is spent on the 3 add_tag lines here:

    void send_to(io::mcbp_session session)

    {

        if (!handler_ || !span_) {

            return;

        }

        session_ = std::move(session);

        span_->add_tag(tracing::attributes::remote_socket, session_->remote_address());

        span_->add_tag(tracing::attributes::local_socket, session_->local_address());

        span_->add_tag(tracing::attributes::local_id, session_->id());

        send();

    }

This comes from a mix of fmt::format and threshold_logging_span::add_tag.

Similarly in map_and_send, 0.63% TT is spent on:

cmd->last_dispatched_from_ = session->local_address();

due to the fmt::format cost.

Suggestions

Suggestion 1: The Java SDK has an optimisation where if it's a basic span only created for ThresholdLogging purposes (where spans aren't used), it'll skip assigning tags.

Suggestion 2: Maybe the session's local and remote addresses can be cached?