cmake: indicate single precision floating-point for Cortex-M33

ARM Cortex-M33 may implement an optional Floating Point Unit
(FPU) supporting single-precision arithmetic. This commit
modifies the respective GCC_M_CPU flag to comply with Cortex-M33.

Signed-off-by: Ioannis Glaropoulos <Ioannis.Glaropoulos@nordicsemi.no>