[orielhaim]

Adds optional portable-atomic support for targets without native atomic RMW operations, including Cortex-M0/M0+ (RP2040) and RISC-V

These targets don't support operations like compare_exchange and fetch_add, causing compilation errors

With the feature enabled, the atomic types come from portable-atomic instead of core, with critical-section support where needed

Default and loom builds remain unchanged