[webgpu] Add CeilDiv into webgpu utils

[daijh]

Description

The CeilDiv function, which is heavily used by operations to determine dispatch sizes, was duplicated in multiple files. This PR moves the implementation into the common WebGPU utils to ensure a single source and reduce code duplication.

Motivation and Context

See above.

[daijh]

@guschmue @fs-eire @qjia7 PTAL.

[qjia7]

nit: why not use WORKGROUP_SIZE?
In line 134, please help correct it to uint32_t output_size = onnxruntime::narrow<uint32_t>(input_shape.Size());

[daijh]

The reasons:

1. Specify workgroup_size=64 explicitly improve readability.
2. WORKGROUP_SIZE is defined as SafeInt<uint32_t>, it's not easy to cast back into uint32_t which is required by CeilDiv.

[qjia7]

It seems that it's not necessary to use SafeInt<uint32_t> for WORKGROUP_SIZE. WORKGROUP_SIZE is just a default value. If you are using 64 as the workgroup size, you can skip to call program.SetWorkgroupSize(64u); unless you are using a different value. Currently, a lot of ops are using the default WORKGROUP_SIZE. If you search, there should be many 'SetDispatchGroupSize((XXX + WORKGROUP_SIZE - 1) / WORKGROUP_SIZE). @fs-eire Do you think it's better to change the type of WORKGROUP_SIZE to uint32_t from SafeInt<uint32_t>? Or in all places, we explicitly specify the workgroup size.

[daijh]

I personally prefer to explicitly specify the workgroup_size. This practice improves code readability and saves time for developers read the source first time, as it eliminates the need to navigate to another file to figure out the default WORKGROUP_SIZE;

I have no objection if someone wishes to retain the implicit style.

[guschmue]

/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 4 pipeline(s).

[tianleiwu]

remove empty line here.

[daijh]

Thanks for identifying it.