[PATCH v4 13/21] math: Remove the SVID error handling from atan2f

[Wilco Dijkstra]

Hi Adhemerval,

Are we sure there are no other paths that result in underflow
to zero? Like eg. return 0.0 * sgn[uy >> 31]; case or when 
calling cr_atan2f_tiny?

Cheers,
Wilco