This change brings concurrent multi-request execution to test_dflash
on a single GPU. It is internally one cohesive unit but can be split
into four conceptual pieces if a smaller review is preferred:

1. Multi TargetCache slots
   - CLI: --target-cache-slots=N (alias --cache-slots=N)
   - prefix `SLOT <id>` routes commands to a specific slot
   - DaemonSlotState + RAII ActiveDaemonSlot for safe switching
   - LIST_TARGET_CACHE_SLOTS for introspection
   - all slots share target/draft weights; only KV/SSM/scratch is
     per-slot
   - create_target_cache gains an `n_seqs` parameter so a single
     cache can be allocated batched up front

2. Tagged stream protocol (opt-in)
   - --stream-tagged emits frames `[-2, request_id, token]` instead
     of bare int32 tokens; sentinels `-4` (CONTINUE), `-1` (DONE)
   - parser recognises `REQ <id>` / `REQUEST <id>` headers
   - legacy bare-int32 streaming is unchanged when the flag is off
   - this lets a client demux multiple concurrent requests over the
     same stdout

3. Native quantum scheduler
   - dispatch table for REQ/SLOT/START, SCHED_STEP, SCHED_DRAIN,
     LIST_REQUESTS
   - cursor-based fair round-robin between admitted requests
   - non-blocking reader thread admits new requests during a drain
   - PendingQuantum{slot, req, epoch, n_gen} carries the unit of work
   - CONTINUE / CONT resumes a slot without re-prefilling
   - REQ <id> CANCEL invalidates a request and bumps the slot epoch
     so a stale CONTINUE is rejected; RESTORE_CHAIN / legacy generate
     refuse to overwrite a slot that is owned by an active scheduler
     request

4. Fused batched target step (CUDA path)
   - new commands: SCHED_BATCH_PEEK, SCHED_BATCH_PROBE,
     SCHED_BATCH_TARGET_TAIL, SCHED_BATCH_TARGET_STEP,
     SCHED_BATCH_DRAIN
   - QwenGraphInputs gains `n_seqs`; build_delta_net_block accepts
     n_seqs > 1
   - target_feat is allocated as [5*hidden, target_feat_cap, n_seqs]
     when batched and the chain forwards capture features per-seq
   - batch_probe_compare_ok smoke shows mismatches=0 vs the
     single-seq path; SCHED_BATCH_TARGET_TAIL commits two completed
     pending quanta in 29.26 ms; SCHED_BATCH_TARGET_STEP commits the
     next batched step in 29.57 ms; SCHED_BATCH_DRAIN completes
     req12/req13 with two batched steps each
   - rollback for partially accepted draft tokens, multi-token verify
     and parent-id propagation in the batched path are noted as
     follow-ups; today the batched step accepts the cleanest case
     and falls back to single-seq when needed

Validation (single GPU1 RTX 6000 Ada sm_89, Heretic Q4_K_M target +
Q8 GGUF or FP16 safetensors drafter, FA_WINDOW=0, KV q4_0/q4_0):

- Two concurrent requests:
    REQ 4 START SLOT 0 quantum=2
    REQ 5 START SLOT 1 quantum=2
    SCHED_DRAIN closes both clean.
    slot 0: 18.41 tok/s, slot 1: 22.50 tok/s
- Mid-drain admission of REQ 6 succeeds; CONTINUE on slot 0 resumes
  without re-prefill.
- batch_probe_compare_ok mismatches=0 over a 2-seq probe.
- batch_tail_commit count=2 ms=29.26.
- batch_step_commit ms=29.57 followed by SCHED_DRAIN reverts cleanly
  back to the DFlash single-seq path.

Compatibility:
- All new behaviour is opt-in. Default invocation of test_dflash
  with no scheduler flags keeps the legacy single-request path.
- Tagged stream is gated behind --stream-tagged.
- Multi-slot is gated behind --target-cache-slots=N (default N=1).
- Batched target step is reached only via the SCHED_BATCH_* command
  family; legacy SCHED_STEP keeps using the single-seq path.
- Hot-loop diagnostic logs (sync_us / step_debug) are now gated
  behind DFLASH27B_TIMING_DEBUG / DFLASH27B_STEP_DEBUG so the
  default path is unchanged.

Verification vs existing community PRs:
- No prior art in lucebox-hub for the SCHED_BATCH_* protocol or for
  a native C++ quantum scheduler with REQ/SLOT/CONTINUE/CANCEL +
  epoch hardening. Checked against PR Luce-Org#39 (CUDA graph reuse) and
  PR Luce-Org#62 (split target/draft StepGraphs); both reuse / split graphs
  but neither exposes a multi-request slot protocol.
- No upstream collision found for tagged stream framing or
  --target-cache-slots.

Happy to split this into four sequential PRs (slots / tagged stream /
quantum scheduler / batched target step) if a smaller-grained review
is preferred — let me know.

Author: Javier Pazo <xabicasa@gmail.com>