描述 Description

请提供此 Pull Request 的高级别说明。
Added a pass that converts tensors that have non power of two dimensions into tensors with power of two dimensions by padding them.
Supports the following ops:

• arith.constant ops
• tensor.insert_slice, tensor.extract_slice, tensor.insert, tensor.extract ops that are supported by RewriteSliceOpToTriton
• scf if/for ops
• triton advance, atomic_rmw, broadcast, dot, expand_dims, histogram, load, make range, make tensor ptr, reduce, reshape, scan, store, splat, transpose ops
• Elementwise ops (such as arith.addi, etc)

Also modified dynamic offset insert_slice lowering in RewriteSliceOpToTriton to support offsets that are not multiples of the insert tensor's insert axis dimension, and added a fallback to using the dynamic slice op lowering if the static offset is not a multiple of the small tensor axis dimension.

类型 Category

• Bug修复 Bugfix
• 新功能 New feature
• 代码风格更新（格式化，局部变量）Style changes
• 重构（即不是新增功能，也不是修改bug的代码变动）Refactor
• 构建过程或辅助工具的变动 Build-related
• 文档内容更新 Docs

Checklist

• 我的代码遵循这个项目的代码风格 My code follows the project's code style
• 我已经自己测试过我的代码 I have self-tested my code
• 我已经更新了相应的文档 I have updated the relevant documentation
• 我已经根据需要更新了对应的变更日志 I have updated the changelog as needed
• 我已经在标题中正确使用了类型标签（例如：feat:, fix:）I have correctly used type labels in the title (e.g., feat:, fix:)