embassy-mspm0: add CPU accelerated float division function

[bespsm]

• mspm0-mathacl: add division example
• mspm0-mathacl: add CPU accelerated float division function
• using internal Mathacl Qval type for providing length of fractional part

[bespsm]

@i509VCB I added division function and used it it in SINCOS

[i509VCB]

I'm not sure how I feel about doing all the calculations and then convert to I16F16 internally, which can't match extreme ends of real f32 math. The error tolerance makes me question this more.

I do think we should probably make mathacl assume types from fixed point for calculations. TI implemented iqmathlib using the accelerator, so it is clearly intended for that.

It makes more sense that the user understands what is going on by making them explicitly convert to/from f32.

I'm not going to say no given you are doing the work here, but maybe I'll ask a bit on matrix because the cordic on stm32 is similar in some ways.

[i509VCB]

From what I can tell, the MathACL does:

• i32
• u32
• u16f16
• i16f16

For DIV anything from i1f31 to i32 and vice versa for signed?