feat: ESP-IDF v6.0 support

[Sped0n]

Change Overview

• Add ESP-IDF version-aware component dependencies in CMakeLists.txt.
• Keep ESP-IDF v5.x using the legacy driver dependency.
• Use split ESP-IDF v6 driver components such as esp_driver_gpio, esp_driver_i2c, esp_driver_spi, esp_driver_uart, esp_driver_rmt, and related dependencies.
• Build the component as C++17 for native ESP-IDF builds. (m5_utility/murmurhash3.hpp:140:30: error: space between quotes and suffix is deprecated in C++23 [-Werror=deprecated-literal-operator])
• Fix ESP-IDF v6 GPIO compatibility by including the needed GPIO register header.
• Guard use of the removed io_loop_back GPIO configuration field so it is only used before ESP-IDF v6.
• Include FreeRTOS explicitly where portMAX_DELAY is used by the SPI adapter.

Testing

Tested with a temporary native ESP-IDF project targeting esp32c6.

Passed:

• idf.py set-target esp32c6 build with ESP-IDF v5.5.3
• idf.py set-target esp32c6 build with ESP-IDF v6.0

[GOB52]

Thanks @Sped0n for the IDF v6 support — much appreciated!

I'd like to merge this as-is so the v6 build path lands, but I plan to follow up with two adjustments on develop:

1. Remove the project-wide -std=gnu++17 override in CMakeLists.txt. The original motivation (C++23 deprecated-literal-operator warning
   from M5Utility) has been fixed upstream, so M5UnitUnified can return to the IDF default and stop forcing a global flag for all
   components.
2. Add a code comment in adapter_gpio_v2.cpp noting the v6 behavior change of rmt_tx_channel_config_t::flags.io_loop_back. On IDF<6 the
   hardware loopback bit is set; on v6 the field is gone and the software loop_count path still works, but the silent feature loss is worth
   documenting at the call site.