Input: PadRIO -- play without the pod (L4CONTROLS=PAD), Workstream A.1

Ported from RP412: RIOBase split out of the serial RIO (L4RIO.h),
rioPointer is RIOBase* (L4CTRL.h), PAD token -> new PadRIO() speaking
the RIO surface from an XInput pad + keyboard (L4PADRIO/L4PADBINDINGS,
vRIO bindings.txt grammar, hot-plug), KeyLight RGB mirror TU
(BT412KEYLIGHT, /std:c++17 per-file).

BT-side fixes PadRIO forced into the open:
- Both keyboard input bridges (mech4.cpp, mechmppr.cpp BT_KEY_BRIDGE)
  stand down when a RIO device exists -- they overwrote the engine
  controls push every frame. M/X conveniences stay live.
- Mapper attribute chain OFF BY ONE (latent real-pod bug): the DOS
  chain below MechControlsMapper carried two base attributes, WinTesla
  carries one, and AttributeIndexSet::Find is positional -- the .CTL
  stick mapping wrote throttlePosition. Pad slot + binary-locked enum;
  gotcha ledgered (reconstruction-gotchas #11).

Verified: PAD throttle lever ramps + sticks, stick turns with the
authentic speed-vs-turn clamp (61.5 -> 22.0 u/s), mech drives; keyboard
fallback intact (BT_FORCE_THROTTLE harness). New diags: BT_CTRLMAP_LOG,
BT_STICK_LOG. (Phase 2 of docs/BT412-ROADMAP.md)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

context/reconstruction-gotchas.md

@@ -206,6 +206,19 @@ as `AttributeIndexSet::Find` crashing in `WeaponCluster::WeaponCluster("PercentD
 (mechweap.cpp now static_asserts its pad base against `PoweredSubsystem::NextAttributeID`).
 [T2]
 
+**The inverse trap — SHORT parent chain (BT412, 2026-07-14):** `Find(ID)` is positional
+(`attributeIndex[ID-1]`), so when the WinTesla parent chain carries FEWER attributes than the
+DOS binary's did, every child attribute resolves one slot EARLY relative to a 1995 stream's
+numeric IDs — no crash, values silently land in the WRONG MEMBER.  Proven live the first time
+a real device drove `MechControlsMapper`: the binary's `.CTL` control-mapping stream uses
+stick=3/throttle=4 (the DOS Subsystem chain had 2 base attrs), the WinTesla chain has 1
+(`SimulationState`), so the stick mapping wrote `throttlePosition` and the throttle mapping
+wrote `pedalsPosition`.  Only the numeric-ID path (CreateStreamedMappings / `.CTL`) is
+affected — gauge databinding resolves by NAME and never saw it.  Fix: a named pad slot at the
+front of `mechmppr.cpp AttributePointers[]` + the enum locked to the binary's values.
+Diagnostic: `BT_CTRLMAP_LOG=1` dumps each streamed mapping's resolved pointer — compare against
+`&mapper->member` (the mech4 `[mppr]` log prints `&stick`/`&thr`). [T2]
+
 ## 12. Frame-pacing trap — the binary assumes a LOCKED 60 fps (task #11)
 
 The 1995 pod ran frame-locked; reconstructed per-frame logic can carry HIDDEN frame-rate