Disable b_fs_zyx_fsv16 not to use cldnn ref kernel (#32235)
### Details:
- b_fs_zyx_fsv16 cldnn ref-kernel is enabled for dynamic shape by
PR#31221. This caused perf regression by using ref-kernel which
previously had used onednn kernel.
target model unet3d_mlperf
- Before PR#31221 : throughput 1.8 FPS
- Wtih PR#31221 : throughput 0.59 FPS
#### The code and line that caused this issue
- Disabled b_fs_zyx_fsv16 in reorder impl : reorder_impls.cpp
#### Reproduction step and snapshot
- Can be reproduced by benchmark_app
`benchmark_app -m unet3d_mlperf\onnx\onnx\FP16\1\ov\unet3d_mlperf.xml -d
GPU -nstreams 2 -nireq 4 -t 10 -b 1 -hint none -data_shape
input[1,4,224,224,160] -shape input[?,4,?,?,160]`
#### Checklist
- [ ] Is it a proper fix? A new ticket is needed to enable this format
with optimized kernel.
- [ ] Did you include test case for this fix, if necessary?
- [x] Did you review existing test that can be extended to cover this
scenario?
### Tickets:
- CVS-173587
Signed-off-by: Min, Byungil <byungil.min@intel.com>
