fix: remove unused GPS pins and GPIO5 role conflict on M5Stack Unit C6L

[jirogit]

Summary
Removes 3 pin definitions from the M5Stack Unit C6L build config:

• GPS_RX=4 / GPS_TX=5: Unit C6L has no GPS hardware, so these are dead definitions.
• SX126X_RXEN=5: shares GPIO5 with GPS_TX=5 above. Note this was not itself dead —
  RadioLib's setRfSwitchPins()/setRfSwitchState() actively drives GPIO5 on every TX/RX
  transition, independent of SX126X_DIO2_AS_RF_SWITCH=true (which switches the antenna via
  the SX1262's internal DIO2, not an MCU GPIO). Removing this line leaves DIO2-based switching
  as the sole antenna switch mechanism on this board.

This is independent of the crash-loop fix in #3088 (TCXO/I2C/PI4IO/TX_LED). During bisection
of that issue, these 3 lines were tested both present and removed with no observed effect on
the crash loop either way, so they're split out here as a separate, unrelated cleanup rather
than bundled into that fix.

Testing

• pio run -e M5Stack_Unit_C6L_companion_radio_usb builds clean.
• TX/RX functionality verified on hardware after removing SX126X_RXEN: sent and
  received messages successfully round-trip against a second node. No functional
  regression observed. GPS_RX/GPS_TX removal has no functional impact (confirmed unused).

Related to #2229 (split out from #3088 during investigation).

[recrof]

If we don't want to use GPIO4 and GPIO5 with GPS, I guess it's better to have it connected to i2c bus? those 2 pins are connected to HY2.0-4P on the bottom.

[jirogit]

Good catch, and it turns out this PR is actually a precondition for that idea.
Traced it through RadioLib: SX126X_RXEN=5 wasn't dead — it was actively
toggling GPIO5 (HIGH/LOW) on every TX/RX transition via setRfSwitchState().
If we left that in place while also wiring G4/G5 as I2C, RadioLib would be
fighting the I2C peripheral for control of the SCL pin on every packet.

MeshCore already has a generic pattern for a second I2C bus
(ENV_PIN_SDA/ENV_PIN_SCL in EnvironmentSensorManager.cpp, used by several
existing boards), so once this PR lands, enabling HY2.0-4P as I2C should
just be -D ENV_PIN_SDA=4 -D ENV_PIN_SCL=5 in platformio.ini — no new
C++ needed. One open question: on the C6, the second I2C bus appears to
map to the LP domain controller rather than the standard HP I2C, so it
needs real hardware verification before I'd call it confirmed. Happy to
open a follow-up issue for that if you want to test it on your unit.