The intrinsic brings regression as below:

## tested on K230-CanMV

with -XX:+UseRVV
o.o.b.j.lang.Integers.compress N/A 500 avgt 5 12364.805 ± 3195.681 ns/op
o.o.b.j.lang.Integers.expand N/A 500 avgt 5 15762.576 ± 4010.350 ns/op
o.o.b.j.lang.Longs.compress N/A 500 avgt 5 27292.340 ± 6964.587 ns/op
o.o.b.j.lang.Longs.expand N/A 500 avgt 5 39925.951 ± 8590.752 ns/op

with -XX:-UseRVV
o.o.b.j.lang.Integers.compress N/A 500 avgt 5 10099.840 ± 2359.345 ns/op
o.o.b.j.lang.Integers.expand N/A 500 avgt 5 4949.358 ± 1268.437 ns/op
o.o.b.j.lang.Longs.compress N/A 500 avgt 5 5453.805 ± 1401.250 ns/op
o.o.b.j.lang.Longs.expand N/A 500 avgt 5 5457.328 ± 1347.508 ns/op

The reason of regression could be: the kernel on K230-CanMV is only 5.10 which is a back port of vector support from 6.5+, so the implementation could be sub-optimal, this is also observed on other benchmark tests.

options of next steps:
1. wait for other hardware to further decide the performance.
2. remove these instrinsic implementation