Coolant Flush end-to-end (Gitea #7): InjectCoolant id-4 handler + drain + FLUSH.PFX cloud + 'H' key

The playtest report (sound plays, no gauge drop, no cloud) traced to the
Reservoir's InjectCoolant chain being dead in three places:

- The handler was never REGISTERED: the binary's Reservoir handler table
  @0x50e680 has one entry {4, "InjectCoolant", @4aee70}; added
  Reservoir::GetMessageHandlers + the press/release handler (press starts
  the flush gated on coolantLevel@0x12C > 0 -- the old body misread +0x12C
  as currentTemperature; release stops it; novice lockout via FUN_004ac9c8).

- Reservoir::InjectCoolant (@4aefa4, 1019 bytes) was an empty stub -- the
  drain the coolant gauge reads.  Reconstructed in full: work list =
  condenser/weapon/heatable chains (+0x7cc/+0x7bc/+0x7ac, roster-walk
  emulation with the binary's duplicate-visit weighting; HeatSink-filtered
  [T2 guarded]) + the linked master sink; per sink with flowScale != 0 move
  squirtMass x flowScale x dt (clamped to the tank / sink capacity) and
  credit pendingHeat with the negative carried-heat delta capped at
  sinkMass x reservoir startingTemperature -- the set%-biased flush of the
  manual (p24), riding the existing heat model.

- Two latent ctor decode bugs surfaced and fixed: the master gate @4af408
  (and @4aeb40 HeatWatcher, swept) reads the OWNER MECH's simulationFlags
  (*(param_2+0x28)), not the resource's subsystemFlags (the misread left
  the CoolantSimulation Performance unregistered); and the capacity scale
  FILDs the bank's INTEGER HeatSinkCount ((float10)*(int*)(link+0x1d0)) --
  the float reinterpret gave ~1e-44 -> a permanently empty tank.
  Capacity = 0.05 x heatSinkCount x streamed CoolantCapacity (BLH: 6.0).

Visual: the binary's mode-1 coolant-effect renderable (FUN_00456a68, built
for classID 0xBC0 on "ReservoirState", part_014.c:5439; tick @part_007.c:
8780) starts psfx 19 = FLUSH.PFX ("Coolant flush", BTDPL.INI) when the
state changes to 1 -- BTSpawnFlushCloud (mech4.cpp) spawns it on the same
alarm edge as an attached emitter at torso height.

Input: new CONTROLS.MAP action "Flush" on 'H' (HELD; press+release both
dispatch, the held-button payload).  Diags: BT_FLUSH_LOG, BT_FLUSH_TEST.

Verified live (FOGDAY): [flush] Reservoir level 6 -> 0.13 -> 0 over ~0.6 s
held (= the manual's 3-4 punches to empty), the coolant vertBar source
Reservoir/CoolantMass drains, and the bluish condensation cloud rises from
the mech (scratchpad/flush_cloud.png vs flush_before.png).

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

content/CONTROLS.MAP

@@ -31,7 +31,7 @@
 #           LeftTrigger RightTrigger (as digital, >= 50% pressed)
 # <src>     LeftStickX LeftStickY RightStickX RightStickY
 #           LeftTrigger RightTrigger
-# action    ViewToggle LookBehind AllStop ModeCycle Valve ConfigHold
+# action    ViewToggle LookBehind AllStop ModeCycle Valve Flush ConfigHold
 #           Generator1-4 Reconnect
 # pckey     sends an authentic 1995 PC-keyboard hotkey (the in-cockpit
 #           dispatcher): '+'/'-' = target zoom, 'w'..'o' = pilot select,
@@ -73,6 +73,7 @@ key D4    button 0x45          # Pinky (4th group)
 # Systems.
 key G     action ConfigHold    # hold + fire keys = regroup the weapon
 key C     action Valve         # coolant valve cycle
+key H     action Flush         # coolant flush (HOLD -- dumps reservoir coolant)
 key M     action ModeCycle     # Basic -> Standard -> Veteran controls
 key V     action ViewToggle    # cockpit <-> chase camera (dev)
 key B     action LookBehind    # HOLD = the pod's rear-view button