Use `u32` or `i32` consistently for ByteView length

[alamb]

We found some inconsistency in the arrow-rs codebase

Some places use *view as i32 and some places use *view as u32. We use a consistent practice and document it.

I think most other places use u32 (rather than i32) -- for example

https://github.com/apache/arrow-rs/blob/6505d2a354e87e193e00ca7c6ea945de92c7adca/arrow-array/src/array/byte_view_array.rs#L372-L371

https://github.com/apache/arrow-rs/blob/6505d2a354e87e193e00ca7c6ea945de92c7adca/arrow-array/src/array/byte_view_array.rs#L339-L338

Originally posted by @alamb in #9363 (comment)

[Jefffrey]

Related:

• ByteView should be using signed types? #6172