fix(firmware): route SPI2 MISO to GPIO35 so the SD card mounts

[andymai]

Summary

• feat(firmware): dynamic SD-SPI clock rate (400 kHz init -> 25 MHz data) #552 added the per-transaction output-enable flip on GPIO35 (the pin CoreS3 shares between LCD DC and SD MISO) but never connected SPI2's FSPIQ (MISO) input signal to the pin in the GPIO matrix. The peripheral therefore read an unconnected input (0x00), embedded-sdmmc spun in its CMD0 retry loop, and the card init failed with CardInit — the SD never mounted.
• Make GPIO35 a Flex with its input buffer kept live, and route FSPIQ to it via the supported InputSignal::FSPIQ.connect_to(&dc). sd_spi's OE flip still hands the line to the card during reads. Both halves (OE flip + MISO route) are now present; updated the sd_spi module doc and the LcdDisplay type alias accordingly.

Test plan

• just check-firmware — clean (exit 0).
• just clippy-firmware (-D warnings) — clean (exit 0).
• On-device (M5Stack CoreS3, brand-new SD card inserted): flashed --release and watched the defmt boot log:
  • Got response: 0 retries: 51 → 1 (one normal init retry, then success).
  • SD: mount failed (CardInit) — gone.
  • Now: SD: read /sd/STACKCHAN.RON failed (FileNotFound); using defaults (card mounts; config file simply absent on the blank card), runtime store: ... seeding from boot config, ble: bonds: 0 persisted bond(s) registered (was "no SD mounted; bonds disabled").
  • ILI9342C ready + render task: 33 ms tick + boot complete — LCD still renders fine through the Flex DC pin; no panic.

Greptile: please review.

[greptile-apps]

Greptile Summary

This PR fixes a missing half of the GPIO35 dual-purpose pin setup that caused SD card initialisation to fail. #552 added the per-transaction output-enable flip so the SD card could drive MISO, but never wired SPI2's FSPIQ input signal to the pin in the GPIO matrix, so the peripheral always read 0x00 and embedded-sdmmc spun in its CMD0 retry loop.

• Upgrades GPIO35 from Output to Flex, enables its input buffer, and calls InputSignal::FSPIQ.connect_to(&dc) so SPI2's MISO is properly routed to the physical pin — the missing half that caused CardInit failures.
• Updates the LcdDisplay type alias to reflect the Flex<'static> DC pin, and rewrites the sd_spi module doc to explain that both the FSPIQ route and the OE flip are required together.

Confidence Score: 5/5

Safe to merge — the change is a focused, well-understood hardware fix with on-device confirmation that both the LCD and SD card work correctly after it.

The fix is minimal and surgical: one pin type change, one InputSignal routing call, a type alias update, and a doc rewrite. The root cause is clearly explained, the on-device test plan covers both affected peripherals (LCD still renders, SD now mounts), and the Flex / InputSignal::FSPIQ approach is the supported esp-hal surface. No logic in the OE-flip path changed.

No files require special attention.

Important Files Changed

Filename	Overview
crates/stackchan-firmware/src/main.rs	Replaces Output with Flex for GPIO35 (DC/MISO shared pin), enables input buffer, and routes InputSignal::FSPIQ to the pin so SPI2 can sample SD card MISO; updates LcdDisplay type alias accordingly.
crates/stackchan-firmware/src/sd_spi.rs	Doc comment updated to describe the full two-part mechanism: the FSPIQ GPIO matrix route (new, in main.rs) and the per-transaction OE flip (existing here); no logic changes.

_{Reviews (1): Last reviewed commit: "fix(firmware): route SPI2 MISO to GPIO35..." | Re-trigger Greptile}