Rename `MLOperandDescriptor`'s `dimensions` to `shape`

[a-sully]

As discussed in #666:

MLOperandDescriptor takes a dimensions field, while querying this field from an MLOperand is done with the shape() method. As per https://w3ctag.github.io/design-principles/#attribute-reuse, we should re-use terms rather than unnecessarily introducing new vocabulary.

A quick CTRL-F of the spec text shows:

• 139 instances of "dimensions", mostly due to MLOperandDescriptor.dimensions
• 483 instances of "shape", which seems to be used more commonly in prose (e.g. "the shape of the input tensor")

We should decide to use one or the other (and possibly continue this discussion on another issue regardless)

"shape" is the more widely used term, so I propose we align on that. WDYT?

[fdwr]

shape is widely used by other ML frameworks

Seems like a tossup, as "dimensions" is widely used by ML libraries too: XNNPACK dims, ANN dimensions, DML DimensionCount and Sizes, MPS MPSNDArraySizes, CudNN CUDNN_ATTR_TENSOR_DIMENSIONS, OneDNN dnnl_dims_t. 🤷‍♂️

"dimensions" is ambiguous, since it can mean [rank or list of dimensions] ...

Interestingly that ambiguity also applies to a field named cats, as you can have a list of cats (std::vector<Cats> cats = {"Fido", "Garfield", "Heathcliffe"}) or a count of cats (uint32_t cats) depending on vector vs scalar usage. 😺 Really though, we should shore up the specification to consistently use rank when we mean rank, not sometimes loosely refer to "dimensions" to mean rank.

Updated my comment with everyone's pro's/con's statements so far, but I have one other idea on #676 (comment)... 🔍.

[zolkis]

🤔 Have you considered sizes? It avoids the concern with dimensions (list of dimensions vs dimension count), it's equally short as shape and easy to spell, and it avoids the grammatical issues with shape ("... shape excludes information about the object's location, scale, orientation, and reflection... [unlike a] figure [which] is a representation including both shape and size..." [1]).

Using sizes instead of dimensions indeed avoids some confusion, but adds some other... and it's less "standard" (de facto) than shape is. All these terms are overloaded. Since Web NN is targeting framework developers, and the term shape is well known among them (list of tensor component dimension counts, whose length is the rank and whose indices are the axes), I'd slightly prefer shape, but I'd not be confused using sizes, provided we include definitions in the spec.