Add BF16 support to CPU and CUDA Pad kernels (#34062)
### Description
- Enable CPU BF16 Pad for opsets 21–25, including 16-bit constant-value
dispatch.
- Register and instantiate CUDA BF16 kernels, including the rank-2
optimized path.
- Add provider-isolated coverage for 2D bottom padding.
### Motivation and Context
BF16 Pad currently fails session initialization with CPU and CUDA
providers, and CPU-only sessions report no implementation. Supporting it
unblocks BF16 graphs that pad rotary caches without requiring a Concat
workaround.
<!-- START COPILOT CODING AGENT SUFFIX -->
- Fixes #33874
---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: tianleiwu <30328909+tianleiwu@users.noreply.github.com>
Co-authored-by: Tianlei Wu <tlwu@microsoft.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
