webnn: Support block-wise quantize/dequantize

Block-wise expands the scale and zero point for both TFLite quantized
tensor and emulated implementation.

For quantized tensor, resize the constant data to match input shape.

For emulated implementation, TFLite resize_nearest_neighbor[1] kernel
only support 2d, so the decomposition [2] is used that expands each
axis by repeating below steps:
1, Compute the flattened shape before and after the given axis.
2, Resize the dimension of axis using nearest neighbor resampling.
3, Reshape back to input dimensions by block_size dimension.

[1]
https://source.chromium.org/chromium/chromium/src/+/main:third_party/tflite/src/tensorflow/lite/kernels/resize_nearest_neighbor.cc;drc=c7f54c35c6e6a69fd778bd212d9c29431a8e48e3;l=76
[2] https://www.w3.org/TR/webnn/#api-mlgraphbuilder-dequantizelinear

Bug: 377172670
Change-Id: I97fd40ef1eeeb63b9b9a62142b2f839bbe750a8f
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/6841515
Reviewed-by: Reilly Grant <reillyg@chromium.org>
Reviewed-by: ningxin hu <ningxin.hu@intel.com>
Commit-Queue: Junwei Fu <junwei.fu@intel.com>
Cr-Commit-Position: refs/heads/main@{#1502528}