feat(profiler/behave_shell): emit motor.keystroke_cadence

BEHAVE-EXTRACTOR.md Phase B Step B.1.

* SessionContext gains typing_bursts: tuple[tuple[float, ...], ...]
  built by _split_typing_bursts(iats) — splits at gaps > IKI_THINK_MAX_S
  (1.5s) and drops bursts of fewer than 3 IATs. Mirrors prototype's
  _split_into_bursts at BEHAVE/prototype_extractors/shell/extract.py:275.
* _features/motor.py:keystroke_cadence(ctx) emits one Observation
  in {steady, bursty, hunt_and_peck, machine}. Median CV across
  typing bursts; mean IKI < IKI_MACHINE_MAX_S paired with CV <
  CV_MACHINE_MAX → machine. Confidence 0.85/0.70/0.65/0.60 per the
  prototype's calibration history.
* < MIN_INPUTS_FOR_CADENCE inputs or zero typing bursts → skip
  emission. v0.1 emits only the burst-CV variant; the prototype's
  NAIVE session-CV variant is parked for v0.2.
* Calibration grid widened (PHASE_A_PRIMITIVES → PHASE_AB_PRIMITIVES)
  to include motor.keystroke_cadence. Grid green across all five
  shards.

Tests: too-few-inputs → no emit, all-think-pauses → no burst → no
emit, uniform IATs → steady, sub-5ms → machine, mixed-pace → bursty,
extreme bimodal → hunt_and_peck.

decnet/profiler/behave_shell/_ctx.py

@@ -20,6 +20,7 @@ from decnet.profiler.behave_shell._parse import (
     hash_token,
 )
 from decnet.profiler.behave_shell._thresholds import (
+    IKI_THINK_MAX_S,
     PASTE_BURST_MAX_IAT_S,
     PASTE_MIN_CHARS_PER_EVENT,
 )
@@ -47,6 +48,9 @@ class SessionContext:
     inter_cmd_iats: tuple[float, ...] = field(default_factory=tuple)
     output_per_cmd: tuple[int, ...] = field(default_factory=tuple)
 
+    # Step B.1 derivations — typing bursts (IATs split at think-pauses)
+    typing_bursts: tuple[tuple[float, ...], ...] = field(default_factory=tuple)
+
 
 def _detect_paste_bursts(
     inputs: list[AsciinemaEvent],
@@ -102,6 +106,22 @@ def _detect_paste_bursts(
     return tuple(bursts), paste_count
 
 
+def _split_typing_bursts(iats: tuple[float, ...]) -> tuple[tuple[float, ...], ...]:
+    """Split a flat IAT sequence at gaps > IKI_THINK_MAX_S.
+
+    Drops bursts of fewer than 3 IATs — too short to compute a stable
+    CV. Mirrors BEHAVE prototype's ``_split_into_bursts``.
+    """
+    bursts: list[list[float]] = [[]]
+    for x in iats:
+        if x > IKI_THINK_MAX_S:
+            if bursts[-1]:
+                bursts.append([])
+        else:
+            bursts[-1].append(x)
+    return tuple(tuple(b) for b in bursts if len(b) >= 3)
+
+
 def _segment_commands(inputs: list[AsciinemaEvent]) -> tuple[Command, ...]:
     """Walk input events, splitting on ``\\r`` / ``\\n`` into commands.
 
@@ -179,6 +199,7 @@ def build_session_context(
         max(0.0, inputs[i][0] - inputs[i - 1][0]) for i in range(1, len(inputs))
     )
     paste_bursts, paste_count = _detect_paste_bursts(inputs)
+    typing_bursts = _split_typing_bursts(iats)
     commands = _segment_commands(inputs)
     inter_cmd_iats = tuple(
         max(0.0, commands[i + 1].start_ts - commands[i].end_ts)
@@ -204,4 +225,5 @@ def build_session_context(
         commands=commands,
         inter_cmd_iats=inter_cmd_iats,
         output_per_cmd=output_per_cmd,
+        typing_bursts=typing_bursts,
     )