[xnnpack] Support bf16 delegation for fully-connected (#21409)

## Summary

Adds XNNPACK delegation support for **bf16** fully-connected, including
bf16 dynamic-quant (`8da4w`). This lets bf16 models (e.g.
`google/gemma-3-1b-it` exported with `--dtype bfloat16`) lower their
linear layers to XNNPACK instead of falling back to portable.

Changes (delegation only):
- **`partition/config/xnnpack_config.py`** — allow `torch.bfloat16` as a
valid partitioned dtype.
- **`operators/node_visitor.py`** — serialize bf16 tensors as
`xnn_datatype_bf16`.
- **`operators/op_linear.py`** — force the bias to fp32 for bf16 FC
(XNNPACK's bf16 FC is `bf16_bf16_f32`: bf16 activation/weight, fp32
bias).
- **`runtime/XNNCompiler.cpp`** — XNNPACK only provides a
`bf16_bf16_f32` fully-connected (bf16 in → fp32 out). When the
serialized graph asks for a bf16 output (fully-bf16 model), define the
FC with an fp32 intermediate output and append an `xnn_define_convert`
(fp32 → bf16) so the delegate boundary stays bf16.

## Dependency

> **Depends on #21400** (vendored XNNPACK bump to `92a7ad5`) — **merge
that first.**

The bf16 dynamic-quant path relies on the `qd8_bf16_qb4w` subgraph
fully-connected added in google/XNNPACK#10818, which only exists in the
bumped revision. This PR is stacked on #21400, so its diff currently
shows the bump commit as well; once #21400 lands I'll rebase and this PR
will reduce to just the delegation changes above.

## Test plan

Exported `google/gemma-3-1b-it` via optimum-executorch (XNNPACK recipe,
custom SDPA + KV cache):

```
optimum-cli export executorch --model google/gemma-3-1b-it --task text-generation \
  --recipe xnnpack --use_custom_sdpa --use_custom_kv_cache \
  --qlinear 8da4w --qembedding 8w --dtype bfloat16 --output_dir <out>
```

- Lowering succeeds; the FC nodes delegate to XNNPACK as
`qd8_bf16_qb4w`.
- Runtime forward on the resulting `.pte` produces finite `bfloat16`
logits, with argmax matching the fp32 and bf16-no-quant baselines
(argmax == 107 on the smoke input).