[Issue]: semantics of a "pixel"

Describe the issue

Spinning this out from discussions in https://github.com/ome/ngff-spec/pull/118. Currently the term "pixel" and "voxel" are used inconsistently by the spec document.

• Inconsistent terminology.
  The coordinate convention refers to "pixel/voxel", but later just "pixel". Usage should be consistent if "pixel" and "pixel/voxel" denote the same concept.
• Geometric vs value semantics.
  In some places in the spec, "pixel" describes a geometrical entity. In other places, "pixel" describes the value stored at an array index, i.e. an intensity.

For consistency, I recommend:

• remove "pixel/voxel" and just use "pixel" throughout
• define "pixel" as a geometrical entity, prior to the first use of the term. Put this definition early in the spec. There's already a reference to the ITK software guide, which contains a definition of the term "pixel" as an N-dimensional box. This definition is problematic for a few reasons[^1] but it's probably good enough to start with. The spec already contains language about how the value at a pixel should be associated with the pixel center. Use the term "pixel center" consistently to distinguish the pixel region from the point in the center.
• Use the phrase "value at a pixel" or "data at a pixel", etc to refer to array scalars associated with a pixel, i.e. the actual values stored in a Zarr array.

[^1]: The ITK definition of a pixel relies on a well-defined pixel spacing, which means it is only defined for regularly-gridded data. That leaves out a lot of real data!

Review the spec document and the terminology discussion in PR #118, then search for every use of “pixel” and “voxel,” especially around the coordinate convention and the existing ITK reference. Done means defining “pixel” before its first use, using “pixel” consistently for the geometric region, and distinguishing stored array values as the value or data at a pixel.