[vectorapi] Unary operations rint, floor, and ceil

See use-case discussed here:
https://mail.openjdk.org/pipermail/panama-dev/2024-May/020429.html

Add unary operations for float/double vectors for the equivalent Math.rint/floor/ceil. The scalar operations are already intrinsified to vroundpd on x64, and can be auto-vectorized. So it should be relatively easy to add support for vector operations, reusing existing support for vroundpd and adding support for vroundps.