Fix ESP32 repeater boot crash-loop on blank/first-boot flash

[szmidtpiotr]

Summary

Fixes #2506 ("Constant rebooting when the repeater is started on Heltec V4 with MeshCore V1.15.0").

SPIFFS.begin(true) in examples/simple_repeater/main.cpp never checked its return value. On a freshly erased/first-boot flash, the initial SPIFFS mount can fail, and the format-on-mount-fail path built into SPIFFS.begin() can itself fail on the first attempt. The code previously proceeded regardless, leaving fs/IdentityStore/CommonCLI/ClientACL/RegionMap all operating against a non-functional filesystem object. Reads against SPIFFS in that broken state don't fail cleanly — they can return stale/garbage data — which corrupts node prefs/ACL/region state in memory and crashes shortly after boot (observed as StoreProhibited/IllegalInstruction exceptions, with register contents matching raw erased-flash byte patterns, e.g. 0xff/0xff00/0xff0000/0xff000000). The device then reboots into the same broken state, producing a permanent crash-loop that only a full erase-and-reflash without hitting the same race would clear (unreliable in practice).

Fix

Check SPIFFS.begin(true)'s return value. On failure, retry once with an explicit SPIFFS.format() followed by SPIFFS.begin(), and halt() with a clear debug log message if it still fails — instead of silently continuing with a broken filesystem.

This only changes the ESP32 failure path; the working case (mount succeeds first try) is untouched.

Testing

Reproduced and verified the fix on physical hardware — a Heltec V4 (ESP32-S3, 2MB PSRAM / 16MB flash):

• Built heltec_v4_repeater via PlatformIO from this branch.
• Full chip erase_flash + reflash on the unpatched firmware (v1.14.1 through v1.17.1 all reproduce): device crash-loops every boot on the blank flash, matching the symptoms in Constant rebooting when the repeater is started on Heltec V4 with MeshCore V1.15.0 #2506.
• Same erase + reflash with this patch applied: device boots cleanly, generates/prints its Repeater ID, completes MyMesh::begin(), sends the boot advertisement, and runs stably (verified over multiple clean test windows with the serial port in raw/no-echo mode to rule out host-side artifacts).

Test plan

• CI build passes for the heltec_v4_repeater (and other ESP32 repeater/room-server/companion) environments
• Maintainer confirms this doesn't regress the normal (non-blank-flash) boot path