[libc][math] Add float-only versions for `inv_trigf_utils.h`

Currently, the inverse trig utility functions in `libc/src/__support/math/inv_trigf_utils.h` perform all intermediate computations in double precision. This is inefficient for float16 and bfloat16 math functions.

We should provide float-only versions for `inv_trigf_utils.h` so that float16 and bfloat16 functions can use them efficiently.
This is a tracking-issue for the same