Add the missing TPS2121 status pull-up (R108) and correct the power-mux docs

The power mux's status output had no pull-up, which broke the USB current
budget interlock.

TPS2121 ST is an open-drain output; the datasheet lists RST = 6-20 kOhm under
Recommended Operating Conditions, alongside RILM, so it is an external part.
ST is driven low only while IN2 (the J3 external 5 V header) is sourcing OUT,
and is high impedance whenever IN1 (the J7 USB-C input) is sourcing. U10 gate 2
inverts MUX_ST into EXT_SELECTED, and MAIN_EN = USB_3A_OK OR EXT_SELECTED
enables the 15 A core buck. With MUX_ST floating on USB-C power, a drift low
asserts MAIN_EN no matter what the TUSB320 advertised, so the board would try
to bring up the core rail on a 500 mA port. Every other open-drain net here
already has its pull-up (R7/R8, R30, R33/R34, R98/R104/R105/R106); this one was
missed.

R108 (10 k, VCC3V3_AUX -> MUX_ST) is appended at the end of design_data.py
rather than beside the mux: reference designators are assigned in call order,
so inserting it mid-file would renumber every later resistor and invalidate the
routed board. Placed at (53.0, 101.6), 2.4 mm from the U10 gate it holds up,
on a site checked clear of existing copper, and routed with close_remaining.py.

Also corrects POWER_TREE.md, which credited PR1 for giving the external input
priority. PR1 = GND actually *disables* priority mode; CP2 (tied to VCC3V3_AUX)
is what selects IN2. Since CP2 derives from the mux's own output the behaviour
is two-phase - highest-voltage wins at cold start, IN2 unconditionally once the
aux rail is up - so with both inputs connected there can be one seamless
switchover at startup. Documented, with the numbers: both eFuses UV 3.98 V /
OV 5.80 V, ILIM 2.69 A (USB-C) and 4.04 A (header), mux OV 5.69 V / ILIM 4.5 A.
The two inputs are never summed: both eFuses block reverse current and the mux
bonds exactly one input at a time.

board_routed.kicad_sch/.kicad_sym are now gitignored byte copies of the tracked
schematic; kicad-cli needs them beside the board for --schematic-parity, and
deleting them in the previous commit had silently disabled that check.

Verified: ERC 0/0, DRC 0 violations, 0 unconnected, every net one copper group,
schematic parity unchanged at the 402 pre-existing warnings.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012KaTmCCxCEW4awsmXRNT1d

hardware/board-cache.lib

@@ -5181,4 +5181,18 @@ X 2 2 700 0 200 L 30 30 1 1 P
 ENDDRAW
 ENDDEF
 #
+# BOARD_R108
+#
+DEF BOARD_R108 R 0 30 Y Y 1 L N
+F0 "R108" 0 200 50 H V C CNN
+F1 "10K" 0 100 50 H V C CNN
+F2 "Resistor_SMD:R_0402_1005Metric" 0 -100 50 H I C CNN
+F3 "" 0 0 50 H I C CNN
+DRAW
+S -500 150 500 -150 1 1 10 f
+X 1 1 -700 0 200 R 30 30 1 1 P
+X 2 2 700 0 200 L 30 30 1 1 P
+ENDDRAW
+ENDDEF
+#
 #End Library